Modality effects in compounding with English inflectional morphology.
Another explanation for the dissociation between regular and irregular plurals in compounds is that irregular plurals are represented and processed differently from regulars (Pinker, 1999; Pinker & Prince, 1988, 1992). Pinker and Prince's dual-mechanism theory proposes dissociated systems in which the processing of regular morphology is mediated by classic symbolic rules of grammar (e.g. the plural of regular English nouns is formed by attaching the inflectional morpheme [-s] to the stem [N], e.g. rat + [s] = rats). Conversely, irregulars are stored as memorized pairs of words (mouse-mice) in the mental lexicon.
In terms of how the dual-mechanism model might impact upon compounding, Marcus et al. (1995) have argued that as compounds are the product of joining together two stems from the mental lexicon to form one word, irregular plurals may be used in compounds because they are stored, already inflected, as lexical items. However, regular forms may not be included in compounds because they are products of the application of a rule that takes place outside the lexicon, 'on-line' and at a later stage than compounding in the word formation process.
Thus, as with the level-ordering model, the dual-mechanism model predicts that regular plurals should not occur in the non-head position of a compound. Irregular plurals, on the other hand, are licensed by both the dual-mechanism and the level-ordering models to appear optionally within compounds.
Table 1 presents a summary of the compound production studies carried out with native English speakers to date, including a breakdown of the percentage of regular and irregular plurals produced within compounds in these studies. It is clear from Table 1 that the omission of regular plurals from compounds is a robust experimental finding and that irregular plurals may or may not be included in compounds. Thus, these studies seem to provide support for both the dual-mechanism and the level-ordering accounts of compounding.
In some accounts of the dual-mechanism model, however, it is predicted that, in any language, all examples of regular morphology should be processed in one way and all examples of irregular morphology should be treated in another manner. Pinker and Prince (1992) state that
it is an extremely strong prediction that in any language one should find that phenomena in either of these two clusters (i.e. regular and irregular morphology) should be found exclusively in association with one another, never in association with a phenomenon from the other cluster (p. 246).
However, such a clear distinction between the two types of morphology in compounds is not apparent from the studies summarized in Table 1. In fact, the pattern of results across the studies is far from uniform. While regulars were almost always omitted from compounds, it is not true that irregulars were always included in compounds. In some studies, irregular plurals patterned with regular plurals and were omitted from compounds. Other irregulars, however, have been included in compounds.
Only in the earliest compounding study carried out by Gordon (1985) was there complete uniformity in that the 3-5-year-old children tested included 98% of regular nouns in compounds in the singular form and 98% of irregular nouns in compounds in the plural form. At the other extreme, although Lardiere and Schwartz (1997) also found that their participants included all regular nouns in their singular form, unlike Gordon's finding, their adult native English speakers also included singular irregular nouns in 95% of compounds produced. Between these two extremes, age of participants seems to be a potential determinant for whether the compounds produced included irregular plurals. It would seem from Table 1 that as native speakers mature and become more proficient in the use of their native language, they include fewer irregular plurals in compounds. For instance, Nicoladis (2000) reports that 3- and 4-year-old children include the correct irregular plural in 65% of cases in which the children were required to produce compounds using irregular nouns. The 5-6-years-olds with normally developing language ability in the study carried out by van der Lely and Christian (2000) knew the correct irregular plural in an average of 78% of cases, but they only included it in an average of 61.6% of compounds produced. The older children (6-10-year-olds) tested by van der Lely and Christian demonstrated that they knew the correct irregular plural in an average of 73% of cases but only included it in an average of 55% of compounds produced. Van der Lely and Christian's teenaged participants (aged between 14 and 17;4) were able to name all the correct irregular plurals, but they only included irregular plurals in 28.3% of compounds produced. Similarly, the 15 adult native speakers included in Murphy's (2000) study produced irregular plurals in non-head position in 28% of compounds produced. It thus appears that there may be a developmental trend to exclude irregulars as native English speakers get older.
In Gordon's (1985) level-ordering-based explanation of the compounding phenomenon, he claims that adults hardly ever produce compounds containing irregular plurals, a finding which is borne out in these experiments with adults. In fact, the lack of irregular plurals in compounds (e.g. 'toothbrush' never 'teethbrush' and 'mouse-trap' never 'mice-trap') forms the centre of Gordon's argument that an innate language process such as level ordering must mediate compound production in children. Given that adults rarely include irregular plurals within compounds, children could not learn that irregular plurals (and not regular plurals) are possible in compounds from the input to which they are exposed. Irregular plurals are licensed by the level-ordering model to appear optionally within compounds, but what Gordon fails to explain is why children 'take up the option' to include irregulars in compounds while adults do not. Similarly, the dual-mechanism model argues that irregular plurals may appear optionally within compounds; however, it fails to explain why children seem more likely to select the irregular plural from the lexicon but conversely why adults seem more likely to select the singular form.
Thus, although the level-ordering and dual-mechanism models have put forward explanations for the dissociation between the treatment of regular and irregular plurals in compounds, they have yet to explain the variation in the way irregular plurals are treated in compounds by adults and by children. However, before concluding that there is a developmental element to the inclusion of plural morphology in compounds, it is necessary to investigate whether something as relatively straightforward as methodological factors might be causing the differences seen in the various studies. Moreover, if external factors such as presentation or response modality are affecting the number of plurals included in compounds, then an input driven explanation of the treatment of plurals in compounds needs further consideration.
Direct comparisons can be made between all of the studies that tested child participants. Oetting and Rice (1993) and van der Lely and Christian (2000) replicated exactly the methodology adopted by Gordon (1985) in his original experiment. Thus, they all showed the children they tested visual stimuli and then asked the child, 'What could we call someone who eats X'? (where X was the plural already supplied by the child). In all of these studies, the child heard the noun used in the plural form before being asked to supply a compound using that noun. Thus, the child may have been primed to use the noun in the plural form. Nicoladis (2000) tested root compounds and adopted a slightly different methodology. She showed pictures and asked the children to find a name for the compounds depicted. To test an item such as 'flower chairs' she would say 'here are some flowers' (and show a picture of some flowers), then she would show a picture of some chairs and say 'here are some chairs', and then she would show chairs patterned with flowers and ask 'What could we call these?' Thus, while Nicoladis did not use the non-head noun in the plural form in the compound elicitation prompt (i.e. she did not say, 'What could we call these chairs pattered with flowers?'), she did supply the plural before asking the child to produce the compound. This means that in all the compound studies testing children, the plural was supplied to them before they were asked to form the compound and may have primed some of the responses they gave. In every study in which children were tested, participants were required to supply their answers orally.
It may be impractical, however, to make direct comparisons between the results of the investigations carried out by Lardiere and Schwartz (1997), Murphy (2000), and van der Lely and Christian (2000) on teenagers and adults. The type of questioning stimuli used to elicit compounds was similar in all three studies, but the mode of presentation of the stimuli and the mode in which participants were required to respond were not consistent. Lardiere and Schwartz, Murphy and van der Lely and Christian all based their methodologies on Gordon (1985). Van der Lely and Christian followed Gordon's methodology exactly. Their participants were required to produce the plural before the compounding task and, during the compounding task, the plural they had supplied was repeated back to them. They recorded verbal responses. Murphy used only aural, rather than pictorial and aural, stimuli. Specifically, Murphy read out a list of questions such as, 'What do you call a cat that watches mice?' Thus, Murphy mentioned the noun in the plural form before asking her participants to use that noun in a compound. Murphy also elicited written, rather than spoken, responses. Lardiere and Schwartz presented a series of pictures showing cartoon characters performing particular tasks and asked their participants to make up names for the characters depicted. For instance, a character was shown painting its toes and the target compound was toe/toes painter. Thus, Lardiere and Schwartz employed pictorial stimuli and recorded verbal responses and, as such, used the same modalities as Gordon. Crucially, however, they did not elicit the plural form in advance and they did not repeat the plural form back to the participant.
These methodological differences could be exerting their own influence in dictating the kinds of compounds participants produce. Lardiere and Schwartz's participants never heard the plural used and were required to retrieve the name of the picture from their own mental representation. Conversely, in Murphy and van der Lely's study, participants were provided with a plural and required to hold it in memory while a question was asked to facilitate the production of a compound word. Interestingly, Murphy and van der Lely and Christian report quite different results than Lardiere and Schwartz in terms of number of irregulars included in compounds in that both Murphy's and van der Lely and Christian's participants included 28% of irregulars in compounds but Lardiere and Schwartz's only included 4.8% of irregulars in compounds.
Some research has indicated that if participants are able to make use of information provided by the experimenter, then the surface features of that information may be just as, if not more, likely to be encoded than the semantics of that same information (Blaxton, 1989; Jacoby, 1983; Morris, Bransford, & Franks, 1977; Roediger, Weldon, & Challis, 1989; Weldon, 1991). As Murphy and van der Lely and Christian provided the plural form in their questions, there may have been a tendency for their participants to encode the sound of that plural and use it in the production of the compound rather than retrieving their own solution to how that plural should be employed in the target compound. This may have been the case particularly for very infrequently used plurals where the participant was less sure of the correct form to use in a compound.
Evidence that it is easier to produce an irregular plural than an irregular singular when primed with an irregular plural comes from a study by Buck-Gengler, Menn, and Healy (2004). Buck-Gengler et al. asked adult native English speakers to read (i.e. participants were supplied with the plural) sentences of the form 'a jar containing COOKIES is a' and to then supply the compound that would fill the blank. Sentences including both regular and irregular primes were tested, and reaction time and whether a plural was included in the compound were recorded. They found that the longest reaction times were recorded by participants who responded with an irregular singular after having been presented with an irregular plural (i.e. produced mouse bowl when the stimulus was 'a bowl containing mice is a'). Reaction times were more or less identical from participants who produced irregular plurals in response to a plural sentence (i.e. produced mice bowl when the stimulus was 'a bowl containing mice is a') or produced either singular or plural regular nouns in response to a plural sentence (i.e. produced rat bowl or rats bowl when the stimulus was 'a bowl containing rats is a'). Buck-Gengler et al. (2004) argue that it takes longer to produce a singular irregular noun in a compound when the prompt stimulus is plural, because extra time is needed to inhibit the just primed plural.
With respect to response mode, there is evidence from previous research that participants respond differently on tasks depending on the modality in which they are required to supply a response. Providing written responses to lexical access tasks may require different processing systems than those implicated in producing spoken responses to the same task (Bonin, Fayol, & Gombert, 1998).
Hence, we need to conduct an experiment which, as well as testing a larger number of mature native speakers than in previous compounding experiments, also compares presentation and response modalities within a single study. By comparing both presentation and response modalities, it should be possible to unravel whether and which methodological factors are responsible for the inconsistencies in the proportion of irregular plurals included in compounds in the various teenager/adult compounding studies.
This experiment was a mixed design with one within-participants factor--type of noun (regular, irregular) and two between-group factors--mode of presentation (visual or aural) and mode of response (oral or written). The dependent variable was the number of plural nouns of each type that participants included in their compounds. Twenty participants were shown pictorial stimuli and of these, 10 were asked to produce compounds orally and 10 in writing. The remaining 20 participants had the stimuli read out to them and of these, 10 were asked to produce compounds orally and 10 in writing.
Participates were 40 undergraduate students in the Department of Psychology at the University of Hertfordshire who took part in the study in exchange for course credit. All were native English speakers and had been educated in the UK continuously between the ages of 5 and 18 years. Of the participants, 28 were aged between 18 and 24 years; 11 were aged between 25 and 44 years; and one participant was aged between 45 and 60 years. There were 39 female participants and one male participant.
Four mass nouns (rice, water, glass and grass) were used to train participants and familiarize them with the task. The test stimuli consisted of the seven irregular nouns that occur frequently in English and seven semantically matched regular nouns. There are only seven frequently occurring irregular plurals in English and this factor severely limited the number of items that could be tested in this experiment. Appendix A shows the full list of test stimuli used. Frequency counts (Kucera & Francis, 1967) for each of the nouns is shown in Appendix B. An independent samples t test, t(12) = 1.435 p > .05, found no difference in frequency between the group of regular nouns and the group of irregular nouns used in the experiment.
For pictorial presentation, both the training and test nouns were represented by black-and-white line drawings of plural items. The pictures measured 15 cm wide by 13 cm long and were mounted on sheets of A4-sized white laminated card. The pictures were piloted to ensure that they elicited the intended response.
Participants were tested individually in an experimental cubicle at the University of Hertfordshire. A preliminary briefing took place during which participants were told that the experiment would involve putting two separate words together to form a new word. They were informed that they would be asked to make up compound words that described someone performing a particular task. The experimenter gave the participants two examples, So for example, you could call someone who drinks wine a 'wine drinker' and you could call someone who cuts grass a 'grass cutter'.
In the visual conditions, participants were shown picture representations of four training nouns and asked to produce a compound in response to the experimenter's questions. For example, the experimenter showed a picture of rice and asked, 'What do you call someone who boils this?' and the participants were to respond, 'A rice boiler'. In the aural conditions, the experimenter asked the participants, 'What do you call someone who boils rice?' and again participants were to respond, 'A rice boiler'. On the rare occasion that a participant did not produce the appropriate compound, the experimenter provided further examples until the participant understood the form of compound that was required.
Once the participants had completed the training session, they moved on to the test questions that were delivered in exactly the same way. Participants in the oral response conditions were asked to speak clearly into the tape-recorder. Participants in the written response conditions were asked to write their responses on the response sheet with which they had been provided. The order of the 14 test items was randomized for each participant.
Performance of all 40 participants on seven irregular and seven semantically matched regular plurals was compared. Results are reported for both participants and items. We carried out t tests on participant analyses and are denoted by [t.sub.1] (ANOVAs as [F.sub.1]) and items by [t.sub.2] (ANOVAs as [F.sub.2]). Effect sizes are denoted by [[eta].sup.2] values.
Table 2 provides an indication of the extent of pluralization in compounds as it shows the mean percentage of regular and irregular compounds in which singulars were included and the mean percentage of regular and irregular compounds in which plurals were included.
t tests (paired samples for the participants analysis and independent samples for the items analysis) showed that a higher mean percentage of singular nouns was included in regular compounds compared with irregular compounds, [t.sub.1](39) = 9.23, p < .001; [t.sub.2](12) = 6.08, p < .001.
Two further paired-samples t tests showed that there was a reliable difference between the rate at which regular nouns were included in the singular or plural form in compounds, [t.sub.1](39) = -66.57, p < .001; [t.sub.2](6) = -96.17, p < .001. The difference between the rate at which irregular plurals were included in the singular and plural form was not reliable in the participants analysis, [t.sub.1] < 1, but was in the items analysis, [t.sub.2](6) = -2.03, p < 0.05.
The focus of the study was to determine whether there were differences in the number of regular and irregular plurals included in compounds, specifically when different presentation and response modalities were adopted. The mean percentage rates of inclusion (with their standard deviations) of regular and irregular plurals in the two different presentation and response modalities are shown in Table 3.
To test whether irregular plurals were treated differently in different presentation and response modalities, an ANOVA with two between-participants factors--presentation modality (measured at two levels: visual and aural) and response modality (measured at two levels: written and oral) was carried out. A repeated measures ANOVA with two within-items factors: presentation modality (measured at two levels; visual and aural) and response modality (measured at two levels: written and oral) was also conducted. This revealed a reliable main effect of presentation modality, [F.sub.1](1, 36) = 6.749, p < .05, [[eta].sup.2] = 0.158; items [F.sub.2](1, 6) = 9.131, p < .05, [[eta].sup.2] = 0.504. There was no reliable main effect of response modality; [F.sub.1](1, 36) = 3.022, p > .05, [F.sub.2](1, 6) = 3.561, p > .05, or reliable interaction between presentation and response modality; F < 1 for participants and items.
An ANOVA with two between-participants factors--presentation modality (measured at two levels: visual and aural) and response modality (measured at two levels; written and oral)--and a second repeated measures ANOVA with two within-items factors--presentation modality (measured at two levels; visual and aural) and response modality (measured at two levels: written and oral)--were also carried out for regular plurals. The main effect of presentation modality, [F.sub.1](1, 36) = 1, p > .05; response modality, [F.sub.1](1, 36) = 1, p > .05; and the interaction between presentation and response modality, [F.sub.1](1, 36) = 1, p > .05, were not reliable in the between-participants analysis. However, both main effects, [F.sub.2](1, 6) = 6, p < .05, [[eta].sup.2] = .504 for presentation modality; [F.sub.2](1, 6) = 6, p < .05, [[eta].sup.2] = .504 for response modality; and the interaction between presentation and response modality, [F.sub.2](1, 6) = 6, p < .05, [[eta].sup.2] = .504, were reliable in the between-items analysis. However, these differences in the number of regular plurals included in compounds when presentation and response modality were manipulated were due to one participant in the aural presentation, oral response condition, who included four regular plurals in the compounds produced. None of the other 39 participants produced any regular plurals in any of the compounds they produced.
Figure 1 shows the mean percentage of plurals of both types that were included in compounds when presentation modality was manipulated. From Figure 1, it is evident that more irregular plurals were included in compounds when the stimuli were presented aurally (56%) than when they were presented visually (35%). A planned comparison revealed that this difference was reliable, participants [t.sub.1](38) = -2.56, p < .05; items [t.sub.2](6) = -2.83, p < .05.
Figure 1 also illustrates that more regular plurals were included in compounds when stimuli were presented aurally (2.9%) than when they were presented visually 0%). The difference in regular items in the different modalities was not reliable for participants, [t.sub.1](38) = 1, p > .05; but was for items, [t.sub.2](6) = -2.32, p < .05.
[FIGURE 1 OMITTED]
The treatment of specific items in compound formation
From Table 4, a pattern is evident in which participants include the lowest token-frequency irregular plural nouns in compounds twice as often as those with the highest token frequencies, t(2) = -4.44, p < .05. The two irregular plurals with the lowest token frequency, 'mice' and 'geese' (token frequency 9 and 3, respectively; Kucera & Francis, 1967), were included in an average of 65% of compounds. Conversely the high frequency irregulars, 'men', 'women' and 'children' (token frequency 752, 184, and 346, respectively; Kucera & Francis, 1967), were only included in an average of 25% of compounds. By appearing in the singular in the majority of compounds, high-frequency irregular plurals pattern with the high type frequency regular plurals.
The present study tested a larger number of mature native speakers than in previous compounding experiments. In addition, this study compared, for the first time, presentation and response modalities within a single study. In this study, regardless of presentation or response modality adopted, a much higher percentage of irregular plurals were included in compounds than regular plurals. This behavioural dissociation between the treatment of regular and irregular plural morphology in compounds has been reported in many previous compounding experiments testing native speakers (Gordon, 1985; Lardiere & Schwartz, 1997; Murphy, 2000; Nicoladis, 2000; Oetting & Rice, 1993; van der Lely & Christian, 2000).
Varying response modality had no effect on the number of regular or irregular plurals included in compounds. Varying presentation modality affected the inclusion of irregular but not regular plurals in compounds. More irregular plurals were included in compounds in the aural presentation mode where participants were given the irregular plural in the elicitation prompt than in the visual presentation modality where participants never heard the plural used and were required to retrieve the name of the picture from their own mental representation. These findings appear to support previous research that has indicated that if, in performing a task, participants are able to make use of information provided by the experimenter then the surface features of that information are likely to be used by the participant (Blaxton, 1989; Jacoby, 1983; Morris et al., 1977; Roediger et al., 1989, Weldon, 1991). This finding may explain the different results obtained by Murphy (2000) and Lardiere and Schwartz (1997). Murphy used an aural presentation and found that participants included 28% of irregulars in compounds, whereas Lardiere and Schwartz used a visual presentation and found that only 4.8% of irregulars were included in compounds. The direction of the difference observed in the presentation modalities in this study clearly patterns with these earlier studies. However, the situation is complicated by the fact that van der Lely and Christian (2000) obtained similar results to Murphy using a pictorial presentation mode similar to Lardiere and Schwartz. However, van der Lely and Christian elicited the plural from their participants before asking them to produce compounds and as such they may have been primed to use the plural in compounds. Furthermore, van der Lely and Christian tested younger participants than Murphy and, from the analysis of previous studies (see Table 1), age of participant (in addition to modality) seems to influence the number of irregular plurals included in compounds.
Thus, methodological differences may partly explain the fact that irregular plurals have been treated differently in the various studies summarized in Table 1. However, even when controlling for modality effects, mature adult English speakers include fewer irregular plurals in compounds than younger English speakers do. Thus, the dissociation between regular and irregular plurals in compounds is only clearly evident in children. Both the level-ordering model and the dual-mechanism model have yet to explain this apparent developmental aspect to the inclusion of irregular plurals in compounds. In fact, the optional nature of irregulars in compounds does render the dual-mechanism model somewhat unfalsifiable in terms of its relevance to compounding. If adults, like children, take up the option of including irregulars in compounds, then they manifest the characteristic dissociation predicted by the dual-mechanism model. However, if adults do not include irregulars, then there is effectively no dissociation between regulars and irregulars in compounds; however, proponents of the dual-mechanism model still argue that this lack of dissociation is licensed by the dual-mechanism model. Therefore, the dual-mechanism model may not be the best explanation of the dissociation when it does occur.
Input-based explanation of compounding
However, an alternative explanation based on the frequency of occurrence of items in the linguistic input might explain the treatment of both regular and irregular plurals in compounds (Murphy, 2000). Looking first at regular plurals, adult English-speaking participants seem to omit regular plurals from compounds, regardless of the manner in which the stimulus is presented or the modality in which they are asked to supply responses. This robust experimental finding may be a result of fact that regular morphology is governed by a rule-based system. However, it might equally be a result of the fact that regular plurals are far more frequent in the English language than irregular plurals. Regular plurals make up 98% of noun types and 97% of noun tokens (Marcus, 1995). Thus, 98% of all nouns in English add an [-s] sound to make their plural and 97% of all plural usage in English involves the processing of the [-s] sound at the end of a noun. This means that adult English speakers will have had a great deal of practice in using the regular plural morpheme at the end of words. The frequency counts of the Wells corpus have shown that the plural [-s] morpheme is never followed by a second noun. Conversely, the possessive [-'s] morpheme is always followed by a second noun. It might be, therefore, that a noun rarely follows the regular plural [-s] morpheme (i.e. patterns such as '*rat[s] chaser' do not occur) because the pattern noun--morpheme [-s]--noun is reserved for marking possession (such as rat's tail). There is some preliminary evidence for this idea (Hayes, Murphy, Davey, Smith, & Peters, 2002).
However, varying presentation modality did affect the inclusion of irregular plurals in compounds. Again, it may be that irregular plurals are affected by presentation modality in a way that regulars were not because, unlike regulars, they are not mediated by an 'automatically' applied rule. However, the token frequency of individual irregular plurals could be the main factor that influences the way they are treated in compounds. Irregular plurals have a much lower type frequency than regulars and while some irregulars are phonologically similar (e.g. mouse-mice/louse-lice), there is no one dominant phonological pattern within the set of irregular plurals. Furthermore, because irregulars are not formed by the addition of the plural [-s] morpheme, they do not compete with the possessive structure (as mentioned above) and therefore they may be followed by the second noun.
Research manipulating the token frequency of regular and irregular plurals might provide further insight as to why all regulars are treated the same in compounds but irregulars manifest greater variability regarding plural marking in compounds. Ellis and Schmidt (1998) required participants to learn a miniature artificial language (MAL). The regular plural prefix used in this MAL had an overwhelmingly higher token frequency than the individual irregular plural patterns used, although some of the irregulars had very high individual token frequencies (as is the case with the plurals in English; Marcus, 1995). Ellis and Schmidt showed that in the very earliest stages of language learning both regular and irregular morphology were subject to token-frequency effects. With increased exposure to a language, however, token-frequency effects on the regular plurals disappeared due to the power law of practice. The power law of practice states that the amount of improvement shown in the processing of particular items decreases as a function of increasing exposure to those items (Anderson, 1982). This effect illustrates that it is difficult to influence performance as it approaches asymptote by introducing extraneous variables such as changing presentation modality. Thus, in both the MAL and in English, the regular plural affix develops 'high lexical strength' (see Bybee, 1995) and becomes invulnerable to contextual effects. Hence, methodological factors would be unlikely to influence the inclusion of regulars in compounds but could influence irregulars.
As suggested above, well-practised language users (i.e. adult native speakers) learn that the high token-frequency regular plural affix [-s] goes at the end of words and that to include the regular plural within a noun--noun sequence would confuse it with the possessive morpheme. High token-frequency irregulars may also have accrued enough 'lexical strength' (see Bybee, 1995) to 'withstand' the influence of variables such as presentation modality. Thus, like regulars, regardless of context, irregulars also are excluded from appearing in the middle of words such as compounds. While having a considerably higher token frequency than even the highest token-frequency irregular plural, the regular plural affix may pattern with high token-frequency irregulars because of the semantic link of plurality (i.e. that any plural noun, regardless of token frequency, conveys the concept of 'more than oneness'). Lower token-frequency irregular plurals, however, may not have enough 'lexical strength' to withstand these external influences, and thus, in some contexts, language users may be 'tempted' to include them in the middle of words such as compounds. Certainly, in this experiment, when participants were given the plural form, it seems that they were 'primed' to repeat the plural form in more than half of the compounds produced using irregular plurals. Where they had to identify a picture from their own mental representations, they were less likely (in 37% of cases) to include the plural form of the irregular. Indeed, in this experiment, participants included the lowest token-frequency irregular plural nouns in compounds twice as often as those with the highest token frequencies did. The two irregular plurals with the lowest token frequency 'mice' and 'geese' (token frequency 9 and 3, respectively; Kucera & Francis, 1967) were included in an average of 65% of compounds. In contrast, the high frequency irregulars, 'men', 'women' and 'children' (token frequency 752, 184, and 346, respectively; Kucera & Francis, 1967) were only included in an average of 25% of compounds. The difference between the results of the present study (i.e. that 56% of aurally presented irregular compounds contained plurals) and those reported by Murphy (2000; where 28% of aurally presented irregular compounds contained plurals) may be partly explained by the items that were tested. Murphy did not test the low-frequency item 'geese', which was the most frequently included item in this experiment (included in 70% of opportunities).
The difference between the number of irregular plurals included by children and adults in compounds may result from the fact that children have processed fewer plurals than adults have. The children tested in the various compounding experiments may have sufficient experience with regular plurals to learn that the [-s] morpheme goes at the end rather than in the middle of words. Furthermore, they may also have learned that the pattern noun--morpheme [-s]--noun is reserved for marking possession. However, they will have experienced far fewer of any one irregular plural pattern and hence the frequency effect will not have 'kicked in'. Thus, they tend to include all irregulars in compounds. Correspondingly, the French ESL participants tested by Murphy (2000) and the Spanish ESL participants tested by Lardiere (1995), who included a large number of regular plurals in compounds, may not have had enough experience with English regular plurals to learn that they always go at the end rather than in the middle of words. Furthermore, given the fact that possession is not marked with the same morpheme as plurality in their native languages (i.e. in Spanish, 'mother's house' is 'la casa de mama' and, in French, it is 'la maison de ma mere'), they may also have yet to master the competitive relationship between the plural and the possessive morpheme in English.
Influence of individual plural items
Individual irregulars may exert their own influence in compounding. For instance, there is already considerable evidence to show that 'feet' and 'teeth' are frequently included in the plural form (in 26% and 21% respectively of opportunities in this experiment). The inclusion of 'feet' and 'teeth' is perhaps due to a semantic effect in that they are treated like some kind of collective noun (Lardiere & Schwartz, 1997, Murphy 2000). Other research (Haskell, MacDonald, & Seidenberg, 2003) has suggested that the sound of an item may affect its inclusion in compounds. According to Haskell et al. (2003), irregular plurals are only included when they do not sound like regular plurals. Hence, 'feet' would be included, but items such as 'mice' and 'geese' would be omitted. However, 'geese' and 'mice' which Seidenberg et al. argued would be omitted from compounds because they 'sound' like regulars were the most frequently included irregular plurals in this experiment. While these results provide mixed support for Seidenberg et al.'s ideas, the point is that not all irregular plurals are treated the same in the compounding task. Alegre and Gordon (1996) also point out that there are examples of regular plurals occurring in compound words in English (e.g. 'drinks cabinet' 'parks commissioner') and, as such, it seems that the inclusion of plural morphology in compounds is much more of a probabilistic process than that allowed for in dual-mechanism accounts.
Further research is required to investigate the role that token frequency, semantics, and phonetics might play in how a particular irregular is treated in a compound. These factors need to be systematically tested in a cross-sectional study (perhaps using a language which includes more examples of irregular plurals than English) to provide a more precise indication of how items with different token frequencies, semantics and phonetics are treated by different age groups.
In conclusion, the behavioural dissociation between the treatment of regular and irregular plurals in compounds could be mediated by the frequency of the two types of morphology in English. In other words, this compounding phenomenon, rather than being the result of differences in the output of two separate mechanisms, is a consequence of the fact that the token frequency of the regular plural morpheme is far more frequent than the token frequency of any one irregular plural. This frequency factor may work alongside the fact that regular plurals are omitted from compounds because the pattern noun--[-s] morpheme--noun is reserved for marking possession to ensure that regular plurals are never included in English compounds. Irregulars are not formed by the addition of the [-s] morpheme; therefore, they do not compete with the possessive construction and, as such, may be included in compounds. Thus, the dissociation between the treatment of regular and irregular morphology in compounds may also result from the fact that one type of morphology is subject to competition with the possessive morpheme, but the other type is not.
Alegre, M. A., & Gordon, P. (1996). Red rats eater exposes recursion in children's word formation. Cognition, 60, 65-82.
Anderson, J. R. (1982). Acquisition of a cognitive skill. Psychological Review, 89, 369-406.
Blaxton, T. A. (1989). Investigating dissociations among memory measures: Support for a transfer appropriate processing framework. Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 657-668.
Bonin, P., Fayol, M., & Gombert, J. (1998). Experimental study of lexical access in the writing and naming of isolated words. International Journal of Psychology, 33, 269-286.
Buck-Gengler, C. J., Menn, L., & Healy, A. F. (2004). What mice trap tells us about the mental lexicon. Brain and Language, 1-3, 453-464.
Bybee, J. (1995). Regular morphology and the lexicon. Language and Cognitive Processes, 10, 425-455.
Ellis, N.C., & Schmidt, R. (1997). Morphology and longer distance dependencies: Laboratory research illustrating the A is SLA. Studies in second language acquisition, 19, 145-171.
Gordon, P. (1985). Level-ordering in lexical development. Cognition, 21, 73-93.
Haskell, T. R., MacDonald, M. C., & Seidenberg, M. S. (2003). Language learning and innateness: Some implications of compounds research. Cognitive Psychology, 47, 119-163.
Hayes, J. A., Murphy, V. A., Davey, N., Smith, P. M., & Peters, L. (2002). The/s/ morpheme and the compounding phenomenon in English. In W. D. Gray & C. D. Schunn (Eds.), Proceedings of the 24th Annual Conference of the cognitive science society. Mahwah, NJ: Erlbaum.
Jacoby, L. L. (1983). Remembering the data: Analysing interactive processes in reading. Journal of Verbal Learning and Verbal Behavior, 22, 485-508.
Kiparsky, P. (1983). From cyclic phonology to lexical phonology. In H. van der Hulst & N. Smith (Eds.), The structure of phonological representations (pp. 131-175). Dordrecht: Foris.
Kucera, H., & Francis, W. N. (1967). Computational analysis of present-day American English. Providence: Brown University Press.
Lardiere, D. (1995). L2 acquisition of English synthetic compounding is not constrained by level-ordering (and neither probably is L1). Second Language Research, 11, 67-269.
Lardiere, D., & Schwartz, B. (1997). Feature-marking in the 1.2 development of deverbal compounds. Journal of Linguistics, 33, 327-353.
Marcus, G. F. (1995). Children's overregularisation of English Plurals: A quantitative analysis. Journal of Child Language, 22, 447-459.
Marcus, G. F., Brinkmann, U., Clahsen, H., Weise, R., & Pinker, S. (1995). German inflection: The exception that proves the rule. Cognitive Psychology, 29, 189-256.
Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16, 519-533.
Murphy, V. A. (2000). Compounding and the representation of inflectional morphology. Language Learning, 50, 153-197.
Nicoladis, E. (2000). Acquisition of compound nouns: When comprehension does not fully precede production. Paper presented at the Compound Processing Symposium. Vienna, Austria.
Oetting, J. B., & Rice, M. (1993). Plural acquisition in children with specific language impairment. Journal of Speech and Hearing Research, 36, 1236-1248.
Pinker, S. (1999). Words and rules. London: Weidenfeld & Nicholson.
Pinker, S., & Prince, A. (1988). On language and connectionism: Analysis of a Parallel Distributed Processing model of language acquisition. Cognition, 28, 73-193.
Pinker, S., & Prince, A. (1992). Regular and irregular morphology and the psychological status of rules of grammar. In L. A. Sutton, C. Johnson, & R. Shields (Eds.), Proceedings of the 17th Berkeley Linguistics Society, 230-251.
Roediger, H. L., Weldon, M. S., & Challis, B. A. (1989). Explaining dissociations between implicit and explicit measures of retention: A processing account. In H. L. Roediger, III & F. I. M. Craik (Eds.), Varieties of memory and consciousness: Essays in honor of Endel Tulving. Hillsdale, NJ: Erlbaum.
Seidenberg, M.S., & MacDonald, M.C. (1999). A probabilistic constraints approach to language acquisition and processing. Cognitive Science, 23, 569-588.
van der Lely, H. K. J., & Christian, V. (2000). Lexical word formation in children with grammatical SLI: A grammar-specific versus an input-processing deficit? Cognition, 75, 33-63.
Weldon, M. S. (1991). Mechanisms underlying priming on perceptual tasks. Journal of experimental Psychology: Learning, Memory and Cognition, 17, 526-541.
Received 13 July 2004; revised version received 25 February 2005
Appendix A List of words used as stimuli in the experiment Irregular noun Regular noun Verb Children Babies Protects Mice Cats Watches Men Boys Kicks Teeth Bones Breaks Feet Hands Washes Geese Swans Keeps Women Girls Paints Appendix B Frequency of use of nouns (from Kucera & Francis, 1967) Regular plural nouns Noun Plural frequency Percentage use in plural form (a) Hands 285 28 Girls 139 27 Boys 138 25 Bones 20 27 Cats 17 29 Babies 12 13 Swans 1 20 Category mean 87 24 Irregular plural nouns Men 752 26 Children 346 36 Feet 283 44 Women 184 28 Teeth 102 45 Mice 9 31 Geese 3 30 Category mean 239.85 34.28 (a) Percentage use in plural form refers to the proportion of times that the noun is used in the plural form out of all times that the noun is used in singular and plural form in Kucera and Francis (1967).
Jennifer A. Hayes (1*), Pamela M. Smith (2) and Victoria A. Murphy (3)
(1) Psychology Department, University College London, UK
(2) Psychology Department, University of Hertfordshire, UK
(3) Department of Educational Studies, University of Oxford, UK
* Correspondence should be addressed to Jennifer Hayes, Department of Psychology, University College London, Gower Street, London WCIE 6BT, UK (e-mail: firstname.lastname@example.org).
Table 1. Percentage of plurals (of either type) included in compounds in previous compounding studies with native English speakers Participants' mean Number included age (N participants/ Regular Correct irregular Study items) plurals (%) plurals (%) Native English-speaking children with normal language development Gordon (1985) 4;6 (33/16) 2 90 Oetting and 5;0 (19/14) 2 59 Rice (1993) Nicoladis 3;0 to 4;0 (25/16) 2.5 65 (2000) van der Lely 5;2 to 6;8 (12/18) 6.6 61.6 and Christian 6;9 to 7;10 (12/18) 1.6 55 (2000) 14;0 to 17;4 (12/18) 0 28.3 Native English-speaking adults Lardiere and Adults (12/16) 0 4.8 Schwartz (1997) Study Presentation modality Response modality Native English-speaking children with normal language development Gordon (1985) Pictorial and aural Oral Oetting and Pictorial and aural Oral Rice (1993) Nicoladis Pictorial and aural Oral (2000) van der Lely Pictorial and aural Oral and Christian (2000) Native English-speaking adults Lardiere and Pictorial Oral Schwartz (1997) Table 2. Mean percentage of plural and singular nouns in compounds with regular and irregular nouns (standard deviations are shown in brackets) Noun Singular (%) Plural (%) Regulars 98.65 (6.25) 1.35 (6.25) Irregulars 53.93 (27.16) 46.07 (27.16) Table 3. Mean percentage (with their standard deviations) of regular and irregular plurals included in compounds for seven regular and seven irregular plurals in the four conditions tested (i.e. pictorial presentation with written responses, pictorial presentation with oral responses, aural presentation with written responses, and aural presentation with oral responses) Pictorial stimuli Aural stimuli Regular Irregular Regular Irregular plurals plurals plurals plurals Written responses 0 (0) 30 (28) 5.8 (1.8) 45 (29) Oral responses 0 (0) 41 (18) 0 (0) 67 (21) Overall mean for 0 (0) 35.5 (23) 2.9 (0.9) 56 (25) presentation modalities Overall mean for response modalities Regular Irregular plurals plurals Written responses 2.9 (0.9) 37.5 (28.5) Oral responses 0 (0) 54 (19.5) Overall mean for presentation modalities Table 4. Number of regular and irregular plurals included in compounds in this study for seven regular and seven irregular plurals and their frequency count as plurals in Kucera and Francis (1967) Number of times included as a plural in compounds, frequency of use of nouns (from Kucera & Percentage number of Item Francis, 1967) shown in brackets times included as a plural Irregulars Men 11 (752) 27.5 Children 10 (346) 25 Feet 26 (283) 65 Women 11 (184) 27.5 Teeth 21 (102) 52.5 Mice 24 (9) 60 Geese 28 (3) 70 Regulars Boys 1 (138) 2.5 Babies 0 (12) 0 Hands 1 (285) 2.5 Girls 0 (139) 0 Bones 1 (20) 2.5 Cats 1 (17) 2.5 Swans 0 (1) 0
|Printer friendly Cite/link Email Feedback|
|Author:||Hayes, Jennifer A.; Smith, Pamela M.; Murphy, Victoria A.|
|Publication:||British Journal of Psychology|
|Date:||Aug 1, 2005|
|Previous Article:||Mood, dissociation and false memories using the Deese-Roediger-McDermott procedure.|
|Next Article:||Retrieval-induced forgetting is inversely related to everyday cognitive failures.|