The use and productivity of verb morphology in specific language impairment: an examination of Swedish (1).

Hansson, Kristina ; Leonard, Laurence B.

Abstract

In an earlier study, we found that Swedish-speaking children with specific language impairment (SLI) differ from younger normally developing compatriots on only a subset of verb morphemes characteristically problematic for children with this type of disorder. In the present study, this finding was replicated, using a carefully devised set of tasks to elicit attempts at the target morphemes. The children with SLI used present copula forms and regular past-tense inflections with lower percentages than both younger and same-age control children. However, the children with SLI were as proficient as the younger controls in the use of present-tense inflections and irregular past forms. Given the pattern of the children's variability and their overregularization of the inflections, the findings could not be attributed to rote memorization of inflected forms. However, the children with SLI were relatively weak in their ability to apply the inflections to nonce verbs. We discuss the findings in terms of their implications for current characterizations of morphosyntactic deficits in SLI, in particular the agreement-deficit, extended optional-infinitive, and surface accounts.

1. Introduction

In recent years, investigators have sought to understand why children with specific language impairment (SLI) have such serious difficulty with verb morphology, especially in languages such as English and German. The problems seem concentrated in finite verb morphology, that is, in those function words (e.g. copula forms, auxiliary forms) and verb inflections that mark agreement and/or tense. In each of these languages, children with SLI are less likely to use finite verb morphemes in obligatory contexts than are younger normally developing compatriots matched for mean length of utterance (MLU) or age controls when MLU is used as a covariate. Examples of studies on English include Cleave and Rice (1997), Hadley and Rice (1996), Leonard et al. (1997), Loeb and Leonard (1991), Marchman et al. (1999), Oetting and Horohov (1997), Rice and Oetting (1993), and Rice and Wexler (1996). Examples of studies on German include Bartke (1994) and Rice et al. (1997). For an altenative view of what constitutes the main problems of German-speaking children with SLI, see Hamann et al. (1998).

Explanations of these extraordinary difficulties have varied. However, one important clue comes from the finding that English-speaking children with SLI usually err by producing bare stems for finite verb inflections (e.g. cook for cooks) whereas, in German, errors usually involve the production of overt infinitives (or participles) in place of finite inflections (e.g. kochen `to cook' for kocht `cooks'). In English, infinitives are bare stems; hence, the errors in the two languages may be the same, despite their superficial differences.

Hansson and her colleagues (Hansson and Nettelbladt 1995; Hansson et al. 2000) have studied another Germanic language, Swedish, with an eye toward the status of finite verb morphology problems in children with SLI. Using spontaneous speech samples as the source of data, Hansson and her colleagues found that Swedish-speaking children with SLI differed from younger MLU controls on certain finite verb forms but not others. Specifically, the children with SLI had greater difficulty than the MLU controls in the use of copula forms and regular past inflections, but not in the use of present-tense inflections and irregular past forms. This pattern of findings was surprising and deserves closer scrutiny. The goal of the present study was to provide a more detailed analysis of finite verb morphology in Swedish-speaking children with SLI.

The most noteworthy aspect of the Hansson et al. findings is that the observed pattern did not conform closely to the predictions of any of the prevalent accounts of grammatical deficits in SLI. For example, according to the surface account (e.g. Leonard et al. 1997), present-tense inflections should prove no easier than regular past-tense inflections because both inflection types are weak syllable forms that rarely appear in sentence positions in which lengthening occurs. Yet, the children with SLI used present-tense inflections--but not regular past-tense inflections--as proficiently as the younger MLU-matched comparison group.

According to the agreement-deficit account of Clahsen and his colleagues (Clahsen 1989; Clahsen et al. 1997; Clahsen and Hansen 1997), children with SLI have special difficulties with asymmetrical relations between categories. These relations are seen, for example, when verbs must agree with the subject according to features such as person or number. However, Swedish verbs do not employ agreement features; the only finite inflections are those that mark tense. Therefore, the problems that Hansson et al. observed in the Swedish-speaking children with SLI seem to fall outside the grasp of the agreement-deficit account.

One account that could accommodate most aspects of the children's pattern of grammatical morphology is the extended optional-infinitive account of Rice and her colleagues (e.g. Rice et al. 1997; Rice and Wexler 1996; Schutze 1997; Wexler et al. 1998). According to this account, children with SLI remain for a protracted period in a stage of development during which they assume that agreement and/or tense is optional. When the agreement and tense options are not selected, the children select an infinitive. In English, of course, infinitives are bare stems (as in run in Let's watch her run). However, in Swedish, infinitives are marked by overt inflections. Hansson et al. found that the most common error in contexts requiring a past-tense inflection was the production of an infinitive. Two other findings from the Swedish work could also be handled by this account, namely, the findings that both present copula forms and regular past-tense inflections were more problematic for the children with SLI than for the MLU controls. Both of these grammatical-morpheme types involve tense.

However, it is not clear why the children with SLI were as capable as their MLU controls in the use of present-tense inflections and irregular past forms. Both of these morpheme types involve tense, yet were used in over 90% of their obligatory contexts by the Swedish-speaking children with SLI.

It can be seen, then, that if these findings for Swedish reflect the true state of affairs, modifications will have to be made to the existing accounts of grammatical deficits in SLI, or wholly new accounts will have to be developed. The purpose of the present study is to see if such actions will be necessary. Specifically, we take a closer look at the status of these verb forms in the Swedish data by performing additional analyses of the same children's spontaneous speech, and by examining these children's responses to probes specifically designed to assess the use of these verb forms.

Additional analyses of the children's spontaneous speech are needed for a better understanding of the children's accurate use of the verb forms. For example, according to Clahsen and Hansen (1997), accurate use of inflected forms should be the result of the child having memorized these particular forms and the contexts in which they are used. They should not be the result of the child selecting the appropriate inflection and applying it to the verb. As stated by Clahsen and Hansen,

SLI children do not have a general paradigm of person and number inflection. Thus, their grammars do not allow them to generate a corresponding finite form for any given verb. Instead, these children only have a small set of (stored) finite verb forms ... (1997: 148).

A similar memorization assumption is made by Gopnik and her colleagues (Gopnik and Crago 1991; Ullman and Gopnik 1994). Miller and Leonard (1998) have devised a means by which this assumption can be tested; this analysis procedure will be employed here. Its importance to the Swedish work is that if it can be shown that the children with SLI were using, for example, present-tense inflected forms as mere memorized lexical items, we could not conclude that present-tense morphology falls outside the bounds of special difficulties for these children.

The probes designed for the present study offer advantages not available in children's spontaneous speech. First, probes can ensure that the children's ability to apply inflections is assessed with a wide variety of verbs. Even large samples of spontaneous speech do not prevent the possibility that children will produce inflected forms only for verbs of a particular type.

Probes also allow for the introduction of nonce words. Such words are especially helpful in testing children's ability to use inflections productively. That is, if children can apply an inflection to words being heard for the first time, it seems very unlikely that their appropriate use of the inflection in known words is due primarily to rote learning.

2. Method

2.1. Participants

The participants of the study were 42 Swedish-speaking children from the south of Sweden. One group consisted of 14 children meeting the criteria for SLI. These children, eight girls and six boys, were between the ages of 4;3 (years; months) and 5;7. All children in this group scored more than one standard deviation below the mean for their age on the grammatical subtest of the Lund Test of Phonology and Grammar (Holmberg and Stenkvist 1983), a test of expressive language ability. Items on this test examine noun plurals, possessive pronouns, prepositions, negation, and verb inflections, among others. In addition to their low test performance, all of these children had been diagnosed by a speech-language pathologist as exhibiting a language problem, especially in the area of grammar, based on assessment of the children's spontaneous production in combination with test results. The mean length of utterance (MLU) in words of these children ranged from 2.36 to 4.41, based on a spontaneous speech sample of 100 complete, intelligible, nonimitative, nonelliptical utterances. The language-comprehension abilities of the children with SLI were assessed by means of the Language Comprehension Test for Children (Hellquist 1982). This test does not provide standard scores. The children with SLI scored significantly lower on this test (p < 0.05) than their same-age peers participating in this study (see below). However, the clinical impressions of the speech-language pathologists were that the children with SLI had greater difficulty with language production than with language comprehension.

All of the children with SLI passed both a hearing screening and an oral-motor screening (Holmberg and Bergstrom 1996). Based on parental report, no child had been suspected of having frank neurological dysfunction or disturbed social-emotional functioning. All of these children scored within one standard deviation of the mean for their age or higher on the Swedish standardization of the Leiter International Performance Scale (Leissner et al. 1962). IQs ranged from 84 to 122 with a mean of 101.5.

A second group consisted of 14 normally developing children, matched to the children with SLI according to chronological age (hereafter, the ND-A children). These children, nine girls and five boys, were also in the age range of 4;3 to 5;7 and lived in the same communities as the children with SLI. Each child in this group was matched to a child in the SLI group to within two months of age. The selection criteria for this group were the same as for the children in the SLI group with the exception that they performed within normal limits in language as well as nonlanguage abilities. These children's MLUs ranged from 4.23 to 6.49 words.

The remaining 14 participants were a group of younger normally developing children matched to the children with SLI according to MLU (hereafter, the ND-MLU children). This group consisted of eight girls and six boys ranging in age from 2;1 to 3;7. These children resided in the same communities as the children with SLI. The MLUs of these children ranged from 2.04 to 4.21 words. Each child in this group was matched to a child in the SLI group according to MLU to within 0.35 words. As was true for the ND-A children, the children in the ND-MLU group scored within normal limits on both language and nonlanguage measures.

2.2. Grammatical morphemes examined

Four grammatical-morpheme types were examined in this study: (1) the copula in present tense; (2) present-tense inflections; (3) regular past-tense inflections; and (4) irregular past-tense forms. The present copula in Swedish is the present-tense form of the verb vara, that is, ar, and is pronounced [[epsilon]]. It is a weak syllable of brief duration. Unlike in English, it functions only as the copula, not as an auxiliary. Tense is the only finite distinction made in the copula; the same form is used regardless of person and number.

The present-tense inflections investigated were those that are clearly distinguishable from the infinitive, that is, the present tense of verbs with a stem ending with a consonant. In most cases, these were present-tense forms that are formed by the addition of the suffix -er [[??]r] to the stem. For example, the present tense form klipper `cuts' involves an addition of -er to the stem klipp. The latter is used as an imperative (klipp! `cut!'). Importantly, the vowel employed in the present-tense inflection differs from the vowel used in the infinitive form of such verbs (e.g. klippa `to cut'). Thus, even if the child is unsuccessful in producing the [r] of the present-tense inflection, the vowel can serve as a contrastive cue. In addition, present-tense forms are produced with the acute accent whereas infinitive forms are produced with the grave accent. Irrespective of accent, the present-tense inflection appears in word-final weak syllables.

Also included were a few cases where the present-tense form has the same form as the stem. For example, the present-tense form hor `hears' is identical to the stem hor `hear!'. These present-tense forms differ from the infinitive (e.g. hora `to hear') in that the latter requires a syllabic inflection.

The regular past tense is the stem plus the suffix -de [d[??]] for verb stems ending with a vowel or voiced consonant and -te [t[??]] for verb stems ending with a voiceless consonant. Examples include gomde `hid' (stem gom `hide!') and lekte `played' (stem lek `play!'). These forms are quite distinguishable from the corresponding infinitive forms (e.g. gomma `to hide', leka `to play'). Like the infinitive, regular past-tense forms are produced with the grave accent.

Finally, irregular past-tense forms have no suffix but involve a modification of the stem, such as a vowel change (e.g. drack `drank,' stem drick `drink!'). In some cases, the irregular past-tense form also contains a consonant that is not present in the stem (e.g. gick `went', stem gd `go!'). Irregular past-tense forms are readily distinguishable from infinitives (e.g. drack `drank', dricka `to drink'). In addition, there are no Swedish verbs whose irregular past forms are identical to their present forms (in contrast to the English verbs hit and put, for example). For a more thorough description of Swedish verb morphology, see Hansson et al. (2000).

2.3. Procedure

2.3.1. Spontaneous speech. During initial sessions with each child, a battery of language and nonlanguage tests was administered to ensure that the child met the criteria for the respective group. During subsequent sessions, spontaneous data were collected during play with an examiner. The speech samples were audiorecorded and then transcribed. The mean sample size measured in number of complete, intelligible, nonimitative, nonelliptical utterances was 488 for the children with SLI, 492 for the ND-A children and 481 for the ND-MLU children. The transcriptions of the samples were coded for the presence or absence in obligatory contexts of present-copula forms, present-tense inflections, regular pasttense inflections, and irregular past-tense forms using Systematic Analysis of Language Transcripts (SALT; Miller and Chapman 1986). The children's use of these forms in spontaneous speech was reported in Hansson et al. (2000). In the present study, we used the spontaneous-speech data to evaluate the memorization assumption.

2.3.2. Probes. Following the speech samples, the children's use of the same grammatical morphemes was then assessed through specifically designed probe tasks. One task was devoted to present copula, and one task was devoted to present-tense inflections. A third task was designed to elicit both regular past-tense inflections and irregular past forms.

The present-copula task consisted of 12 test items. The child was presented with pictures showing two different persons or objects. The examiner provided a model sentence containing a copula by saying something about one of the persons/objects and then encouraged the child to describe the other, as shown in (1).

 (1) a. Katten ar liten men ... (hasten ar stor)
 `The cat is small but ...' (the horse is big)
 b. Pojken ar darnere men ... (flickan ar daruppe)
 `The boy is downstairs but ...' (the girl is upstairs)

Given the pronunciation of ar as a single vowel [[epsilon]], we ensured that for each item a consonant both immediately preceded and followed the copula form. This phonetic context facilitated our ability to determine the presence or absence of the copula in the child's response. As the task is formulated it would be quite acceptable to respond with an elliptical utterance without the copula. Therefore the child was explicitly given the instruction to describe her/his picture in exactly the same way as the examiner described hers.

Fifteen test items were used to elicit present-tense inflections. Again, a sentence-completion task was employed. The examiner showed the child pictures depicting one or more actions. The examiner described one action in the picture and then asked the child to describe another by giving the subject of that action, as in (2).

 (2) a. Mamma lagar mat och Kalle ... (laser)
 `Mommy cooks and Kalle ...' (reads)
 b. Brandkaren kommer nar huset ... (brinner)
 `The fire brigade comes when the house ...' (burns)

Of the fifteen verbs selected for the task, twelve were of the type taking the -er suffix (as in springer `runs', stem: spring! `run!'). The remaining three were of the type where the present-tense form has the same form as the stem (as in hor `hears', stem: hor! `hear!'). For both types of verb, the present-tense forms are distinguishable from the corresponding infinitive forms (e.g. springer-springa, hor-hora).

In the third probe task, thirteen regular and thirteen irregular past-tense forms were elicited. The procedure was similar to the one used for the elicitation of the present tense. Some of the pictures used were the same as in the present-tense task, which the children had already seen, and the child was reminded `You have seen this picture before. Do you remember what the people did?' as in (3a). For other items, the actions were situated in the past (e.g. last summer) and the examiner provided model sentences containing past-tense forms, as in (3b).

 (3) a. Mamma lagade mat och Kalle ... (laste)
 `Mommy cooked and Kalle ...' (read)
 b. Igar fyllde pappa ar. Nu ska vi se vad de gjorde da.
 Hunden lag och sov och vad gjorde morfar med kaffet, jo
 han ... (drack det)
 `Yesterday it was daddy's birthday. Let's see what they
 did then.
 The dog lay down and slept and what did grandfather do with
 his coffee, he ...' (drank it)

Six of the regular past-tense forms had the -de inflection and seven had the -te inflection. Of the irregular forms, twelve involved a vowel change and one had the same form as the stem (sov `slept').

2.3.3. Nonce items. The present and regular past-tense inflections were also elicited in two tasks where the children were asked to inflect nonce words originally constructed by Andersson and Emenius (1970). Each task consisted of four test items. For present-tense inflections, the children were given the infinitive and imperative forms and then were asked to provide the present tense. An example is shown in (4).

 (4) Den har pojken tycker om att glopa. Jag sager till honom glop!
 Sa vad gor han, han ... (gloper)
 `This boy likes to glopa. I tell him glop!
 So what does he do, he ...'

For the regular past-tense inflections of nonce words, the children were first provided with a present-tense form and were asked to provide the corresponding infinitive form. They were then given the imperative form and were asked to provide the past-tense form, as in (5).

 (5) Den har flickan roper. Vad tycker hon om att gora, hon tycker
 om ... (att ropa)
 Igar sa jag fit henne rop! Vad gjorde hon da, hon ... (ropte)
 `This girl roper. What does she like to do, she likes ...'
 `Yesterday I told her rop! What did she do, she ...'

The request for the children to provide infinitive forms along with present and past-tense forms served as a control to determine if any difficulties on the part of the children with SLI or ND-MLU children might be attributable to the task. It is well known that children do not score as well on nonce items as on corresponding items involving real words (Ratner and Menn 2000). Our inspection of the data from Hansson et al. (2000) revealed that all three groups of children showed appropriate use of infinitive forms in over 95% of obligatory contexts in spontaneous speech. Therefore, we reasoned, if the nonce-item task itself was the problem, scores should be low even when infinitive inflections must be added to nonce words.

2.4. Analysis

2.4.1. Testing the memorization assumption. We used an analysis of verb types to test the assumption that the children's use of inflected forms is attributable to memorization. For present-tense inflections, for example, this assumption suggests that the child will start out using, say, springa `(to) run' in all contexts, because it is the only form she has available. This will result in appropriate use in some contexts (e.g. Kirsten kan springa `Kirsten can run') and inappropriate use in others (e.g. Kirsten springa `Kirsten run'). Once the child has memorized springer `runs', she will use it appropriately (e.g. Kirsten kan springa `Kirsten can run', Kirsten springer `Kirsten runs'). Therefore, we expect that any one verb, in present-tense contexts, will be used always appropriately or always inappropriately, depending on whether or not the child has memorized the appropriate form yet. That is, analysis of the speech of a child who still relies on rote memorization should yield two distinct lists of verbs: verbs consistently used appropriately and those consistently used inappropriately. There would be no basis for expecting a large number of the verbs whose forms alternate between appropriate and inappropriate within the same sample.

For this type of analysis, it was necessary to identify those verb types that were produced in obligatory contexts for the target morpheme at least twice. Each of these verb types was assigned to one of three categories. It was regarded as "always correct" if it always appeared with the appropriate inflection (e.g. for present tense, springer, springer). It was assigned to the "never correct" category if it never appeared with the appropriate inflection (e.g. for present tense, springa, springa). If the verb type appeared with the appropriate inflection in some instances, but not other instances, it was placed in the "sometimes correct" category (e.g. for present tense, springa, springer). It was important to use verb types that were produced at least twice by the child because a verb type used only once could not possibly be in the "sometimes correct" category.

We then determined expected and observed values for each of the three categories. Children were excluded if they never made an error on any verb type. Observed values were computed by tallying the number of verbs types for each child in each category. Expected values were computed in the following manner. First, for each child, the percentage correct use of the inflection type was determined across all tokens of the verb types used at least twice. This figure was then multiplied by the number of verb types used in obligatory contexts for the inflection type to determine the expected "always correct" value. The "never correct" value was then computed by multiplying the percentage incorrect use by the number of verb types used in obligatory contexts. For example, assume that a child produced twelve verb types at least twice in obligatory contexts for present-tense inflections, with a total of thirty tokens. Across these thirty tokens, the child's percentage correct was 60%. By multiplying 0.60 times 12, we obtain the expected number of verb types in the "always correct" category, 7.2. Likewise, by multiplying 0.40 by 12, we obtain the expected number of verb types in the "never correct" category, 4.8.

As the memorization assumption has been stated thus far, no word type will appear in its appropriate form at some times and not at other times. Therefore, the expected value for the "sometimes correct" category should be zero. Using the example above, "always correct" would be 7.2, "never correct" would be 4.8, and "sometimes correct" would be 0. However, this expectation may be too extreme. The assumption that children memorize inflected words can be likened to the memorization of irregular forms such as irregular past in normal language development. Children might initially use throw in contexts requiring threw but will soon learn that threw is the appropriate form for past-event contexts. However, Marcus et al. (1992) proposed that overregularization of past tense (e.g. throwed for threw) occurs when the child fails to retrieve the irregular form from memory; in such instances, the default rule of adding -ed is applied. Marcus et al. found in their study that such retrieval failures occurred about 4.2% of times when a past tense was used. When computation was based on word types, the metric of concern in the present study, rather than word tokens, this figure rose to 10.5% (Marcus et al. 1992: 47). It seems plausible that the same process of retrieval failure might occur in the case of memorized inflected words.

For this reason, we added a step in the computation of expected values. Specifically, after computing the value for the "always correct" category, we calculated 10.5% of this value, subtracted it from the "always correct" category, and placed it in the "sometimes correct" category. Returning to the above example, after multiplying 0.105 by 7.2--yielding 0.756--we subtracted 0.756 from 7.2 and placed it in the "sometimes correct" category. Hence, the final expected values for the child would be 6.44, 4.8, and 0.756 for "always," "never," and "sometimes," respectively.

The final step in the analysis was to add the observed and expected values for all children in the group, to arrive at a total observed and expected value for each of the three categories. For each group and inflection type, the observed distribution was then compared to the expected distribution, using [chi square].

2.4.2. Scoring of probes. For each set of probes, we determined the number of items for which the child provided a response that conformed to the grammatical context established for the item. We then tabulated the number of these items for which the child produced the appropriate verb form. Responses that did not conform to the grammatical context were those such as "I don't know," comments about some aspect of the picture not being tested (e.g. "I see a car"), and comments altogether unrelated to the picture. An important exception was the production of an overregularization of a present-tense inflection (see below) or an overregularization of past in place of the appropriate irregular past forms. These, too, were excluded from the accuracy count. However, we made note of them and describe them in the Results section below. A child's data were included for a particular probe only if there were at least five responses that conformed to the grammatical context (excluding overregularizations), regardless of grammatical accuracy. Because the number of obligatory contexts varied from child to child, percentages correct were used as the comparative measure. These were arc-sine transformed prior to statistical analysis. For each set of probes, we performed an analysis of variance (ANOVA) with group (SLI, ND-MLU, ND-A) as the between-subjects variable. Each ANOVA was followed by post-hoc testing at the 0.05 level to determine which of the three groups differed from one another. Effect sizes (d) for the significant differences were then calculated. All significant differences reported below had d values of 1.14 or greater, which represent very large effect sizes (Borenstein and Cohen 1988; Cohen 1988).

Importantly, many of the ND-MLU children were unwilling to participate in the probe tasks beyond a few items. This was particularly true for the children below the age of 3;0. Six of the seven ND-MLU children below age 3;0 provided too few responses to be included in the analysis for one or more verb morpheme types, whereas all seven of the ND-MLU children above age 3;0 provided a sufficient number of responses to be included in the analysis for all verb morpheme types. It can be seen, then, that the requirement of at least five obligatory contexts for inclusion had a disproportionate effect on the ND-MLU group, as will be shown in the Results.

2.4.3. Scoring of nonce items. The scoring of the nonce items was similar to the scoring of the probe tasks. Specifically, we determined whether the child's response conformed to the grammatical context established in the item and, if so, whether the appropriate inflection was produced. If a child produced a real word instead of a nonce word, it was excluded from analysis. Because there were only four items for present-tense inflections, four items for regular past-tense inflections, and four items for the infinitive forms serving as a control, the data were not subjected to statistical analysis. Instead, we report only the number of children in each group who produced nonce words with the appropriate inflection, and the total frequency of these productions for each group.

The nonce items for present-tense inflections and regular past-tense inflections were presented immediately after their respective real-word probe counterparts. For this reason, the children who did not want to participate beyond a few items on the probe task were not presented with the nonce items. Thus, data will be reported for a smaller number of children from the ND-MLU group than for the SLI and ND-A groups.

2.4.4. Transcription and scoring reliability. Recordings of the probes for three children in each group were independently transcribed and scored by a second individual. Transcription reliability was based on the percentage of items for which the original judge and the second judge had identical transcriptions for the relevant morpheme. Percentages of agreement ranged from 92 to 94 across the morpheme types. However, agreement was somewhat lower for the children with SLI and the ND-MLU children than for the ND-A children. Percentages for these groups across morpheme types were 86, 91, and 99, respectively. No one morpheme type yielded unusually high or low percentages of agreement for any of the three groups of children. Scoring according to accuracy showed high agreement between the two judges. For all morpheme types and all groups of children, the percentage of agreement was 100.

Reliability was also calculated for the children's responses to the nonce items. The responses of the same three children in each group were transcribed and scored by the independent judge. Percentages of agreement for transcription between the original judge and the second judge ranged from 79 for regular past-tense inflections to 100 for present-tense inflections. Agreement was somewhat lower for the ND-MLU children than for the children with SLI or the ND-A children. Percentages were 95, 86, and 94, for the SLI, ND-MLU, and ND-A groups, respectively. The percentage of agreement for accuracy was 100% for all morpheme types and groups of children.

3. Results

3.1. Inflection use as memorization?

The first issue addressed was whether the children's use of present and regular past-tense inflections in spontaneous speech could be interpreted as the result of memorization. For each group and inflection type, the observed distribution across the categories "always correct," "never correct," and "sometimes correct" was compared to the expected distribution through [chi square] analysis. The distributions included in the analyses are provided in Table 1.

Recall that for these analyses, only children showing some degree of error on the inflections were included. For present-tense inflections, eleven of the fourteen children with SLI and ten of the fourteen ND-MLU children were included. The ND-A children were not included due to their near-perfect inflection use. Analysis of present-tense inflection use by the children with SLI revealed an observed distribution that differed significantly from the expected distribution, [chi square] (2) = 17.81, p < 0.001. As can be seen from Table 1, the children with SLI were more inconsistent in their present-tense inflection use with the same verb ("sometimes correct") than expected, even when allowances were made for occasional retrieval errors. Similar findings were obtained for the ND-MLU children, [chi square] (2) = 14.09, p < 0.001.

For regular past-tense inflections, only seven of the fourteen children with SLI produced errors, and only one of the fourteen ND-MLU children showed errors. The ND-A children made no errors. Accordingly, analysis was restricted to the children with SLI. The observed distribution for the children with SLI differed significantly from the expected distribution, [chi square] (2) = 14.02, p < 0.001. As seen in Table 1, the number of verb types in the "sometimes correct" category was larger than a memorization assumption would lead us to expect.

3.2. Probe results

The next set of analyses concerned the children's accuracy on the probes. A summary of the results appears in Table 2. As can be seen in this table, at least thirteen of the fourteen children with SLI and thirteen of the fourteen ND-A children provided a sufficient number of obligatory contexts on the different probes for inclusion. However, only seven to ten children in the ND-MLU group met this criterion.

The children differed in their use of copula forms, F(2, 32) = 12.36, p < 0.001. Post-hoc testing revealed that the children with SLI were significantly less accurate than both the ND-MLU and the ND-A children. For each group, all errors were omissions.

A difference was also found for present tense inflections, F(2, 33) = 4.72, p < 0.02. Post-hoc testing indicated that the children with SLI were significantly less accurate than the ND-A children. No other group differences emerged. All but one error (22 of the 23 errors or 96%) by the children with SLI were productions of an infinitive in place of a present-tense inflection (e.g. springa for springer). The remaining error was the production of a stem (e.g. tand for tander) All five of the errors (100%) of the ND-MLU children were infinitives used in place of present-tense inflections. For the ND-A children, two of the three errors were infinitive productions; the remaining error was the production of a stem.

The present-tense forms of three of the fifteen present-tense inflection items were identical to the stems (e.g. hor). Although the infinitive forms of these verbs (e.g. hora) clearly differ from the present tense and stem forms, it is possible that some of the productions that were scored as correct were in fact attempts to produce the stems. For this reason, we also computed each child's score for the twelve items involving -er only. For these verbs, the present-tense inflected form differs from both the stem and the infinitive form. The children's scores on these twelve items were highly similar to their scores for all fifteen items. The mean percentage correct for the children with SLI was 87.04 (86.71 for all items). For the ND-MLU children, the mean percentage correct was 93.81 (94.50 for all items); for the ND-A children, this percentage was 98.21 (98.50 for all items). Thus, there was no evidence that the scores for all fifteen items were obscuring differences between the groups in the use of present-tense inflections.

All three groups of children produced what can be regarded as overregularizations of present-tense inflections. For example, the verb hor--whose present-tense form and stem are identical--was sometimes produced as horer, with an addition of -er to the stem, as is seen in verbs such as springer and klipper. Five different children with SLI produced a total of six such overregularizations. Two children with SLI produced a total of two overregularizations in which a verb requiring -er in present tense was produced with -ar (e.g. klippar for klipper), a present-tense inflection used when the verb stem ends in a vowel. The ND-MLU children and ND-A children produced fewer overregularizations, and all were of the type klippar for klipper. One ND-MLU child was responsible for the two examples found for this group, and one ND-A child produced a single overregularization.

The ANOVA for regular past-tense inflections also yielded a difference, F(2, 35) = 14.59, p < 0.001. Post-hoc testing revealed that the children with SLI were significantly less accurate than both the ND-MLU and the ND-A children. The latter two groups did not differ from one another. Although it can be seen from Table 2 that the mean percentage correct for the children with SLI was quite low, four of the children in this group scored 100%.

Of the 84 errors committed on regular past-tense items by the children with SLI, 45 or 54% were productions of present-tense inflections (e.g. klipper for kippte). Another 35 errors (42%) were productions of an infinitive (e.g. lasa for laste). The remaining four errors were productions of the stem (e.g. lek for lekte). For the ND-MLU children, nine of the thirteen errors (69%) on regular past-tense items were productions of present-tense inflections. The remaining errors (31%) were productions of the infinitive. The ND-A children produced one error in the form of a present-tense inflection, and one error in the form of a stem.

Irregular past forms also showed a significant difference, F(2, 30) = 15.78, p < 0.001. Post-hoc testing indicated that both the children with SLI and the ND-MLU children were significantly less accurate than the ND-A children. The children with SLI and the ND-MLU children did not differ. One of the thirteen irregular past-tense items required an irregular past form (soy) that was identical to the stem form. We examined the responses to this item to determine whether the children with SLI scored higher on this item than on the remaining items. This was not the case. The percentage correct for this item was 42% for the children with SLI, whereas these children averaged 45% correct across all thirteen items. In contrast, the ND-MLU children showed 100% accuracy on this item but averaged only 66% across all thirteen items. If inclusion of this item produced a distortion in the results, then, it would have to be a distortion in the opposite direction. If anything, inclusion of this item made the ND-MLU children--not the children with SLI--look more proficient than they might actually have been.

For the children with SLI, productions of present-tense inflections constituted the most frequent error (e.g. dricker for drack); 41 of the 63 errors (65%) were of this type. Another 21 errors (33%) were productions of the infinitive (e.g. sjunga for sjong). The remaining error was the production of a stem in place of the irregular past form. Productions of present-tense inflections in place of irregular past forms also represented the most frequent type of error for the ND-MLU children. Twelve of the nineteen errors (63%) were of this type. Three errors (16%) were productions of an infinitive, and four errors (21%) were productions of a stem. The ND-A children produced only four errors for irregular past items. These were one present-tense inflection, two infinitive forms, and one stem.

All three groups of children produced overregularizations of past, such as springde `runned' for sprang `ran'. Ten different children with SLI produced a total of 39 such forms. All seven of the ND-MLU children produced overregularizations, with a total of 23. Even the ND-A children showed an abundance of overregularization in their responses to irregular past items. Eleven ND-A children produced a total of 45 productions of this type.

3.3. Responses to nonce items

The children's responses to the nonce items offered an additional view of their ability to produce language in a productive manner. Table 3 provides the number of children in each group showing such novel use for each inflection type examined. It can be seen from this table that the children with SLI were relatively limited in their ability to apply appropriate present and regular past-tense inflections to the nonce words. Although eleven of these children's responses obligated use of the present-tense inflection, only three of these children actually produced the inflection. As noted earlier, fewer of the ND-MLU children participated in the nonce task. Of the five children whose responses obligated use of the present-tense inflection, four actually produced it. Thus, it appears that the ND-MLU children were more likely to succeed than were the children with SLI. This view receives support from the differences between the two groups in total frequency of use of the present-tense inflection. Considering that there were four nonce items for this inflection, the ND-MLU total of fourteen for five children whose responses required this inflection seems higher than the SLI total of eight for eleven children with SLI whose responses obligated the inflection. All of the children in the ND-A group who produced responses requiring the present-tense inflection were successful in producing the inflection on at least one occasion.

The children with SLI had even more difficulty with the regular past inflection than with the present-tense inflection. As can be seen from Table 3, only two of the seven children with SLI whose responses obligated regular past tense actually produced this inflection. Furthermore, these two children produced the inflection on only one item each. In contrast, all of the ND-MLU children whose responses required regular past tense were able to use this inflection type on one or more occasions. The ND-A children were also quite successful on these items, although one child in this group did not produce the inflection in any obligatory context.

An inspection of the data for infinitive forms in Table 3 suggests that the difficulty on present and regular past inflections by the children with SLI cannot be attributed solely to the task itself. Every child in the SLI group who produced a response requiring an infinitive did, in fact, produce such a form. In addition, although fewer children with SLI had obligatory contexts for infinitives than for either present or regular past-tense inflections, infinitives were nevertheless produced with higher total frequency than the other two inflection types. For the ND-MLU and ND-A groups, performance levels were high on infinitive nonce items.

Although all three groups of children had relative success with the infinitive nonce items, we cannot conclude that task demands played no role. Two observations seem especially noteworthy. First, the children with SLI produced a total of eleven responses to nonce items that differed from the target to such a degree that they could not be scored. Approximately half of these could be scored as correct with a very generous rendering; for example, the production [fiko] might be interpreted as a bona-fide attempt at the present-tense nonce form fliper (compare [fiko] for flipper in English). However, there was insufficient independent evidence to support such an interpretation. Interestingly, the ND-MLU children produced no such responses, and only one response of this type appeared in the ND-A data.

The second observation suggesting difficulty with the nonce items was the fact that real words were often substituted for nonce words. The children with SLI produced a total of 41 responses of this type. Twenty-one such responses were also seen in the data of the ND-MLU children. Such a figure is not inconsequential considering that fewer children in this group participated in the nonce tasks. Fifteen real-word substitutions were seen in the data for the ND-A children. For example, a common response to a present-tense nonce item was gor sa `does like that' accompanied by a pointing or gesturing response. In other cases, the children produced the name of a known action that bore some general resemblance to the action represented by the nonce word. The most frequent example involved a form of the verb `fly' (flyger/flog/flyga `flies/flew/to fly') in place of a nonce verb that referred to a child being lifted from the ground by the motion of a propeller on top of his cap. Responses of this type suggest that adopting a novel word heard only once or twice did not come naturally to the children, especially the children with SLI and the ND-MLU children.

4. Discussion

In the Hansson et al. (2000) study of these children's spontaneous speech, the children with SLI were more limited than the ND-MLU children in the use of present-copula forms and regular past-tense inflections. No differences between these two groups of children were seen for present-tense inflections and irregular past-tense forms. In the present study, we attempted to determine whether this pattern of findings represented an accurate portrayal of the facts. One possibility was that the present-tense inflection abilities of the children with SLI were overestimated by their use of a large number of memorized forms (cf. Gopnik and Crago 1991; Clahsen and Hansen 1997). However, our analysis of the distribution of verbs that were always, never, and sometimes inflected for present tense provided no reason to suspect that the children with SLI were relying on rote memorization as their principal means of using inflected forms.

The same proved true for the children's use of regular past-tense inflections, despite the lower percentages in obligatory contexts seen for these forms.

The spontaneous-speech sample data used by Hansson et al. (2000) might also have given the children with SLI the opportunity to rely principally on those particular verbs whose inflected forms they could control. The children's responses on the probes would therefore be a good gauge of the children's abilities, given that the children's choice of verbs was probably more constrained by the contexts established in each item than might have been the case in spontaneous speech. However, the probe data were very much in line with the spontaneous speech data. The children with SLI showed lower percentages of use of present-copula forms and regular past-tense inflections than the ND-MLU children, but the two groups were similar in their use of present-tense inflections and irregular past-tense forms. This is precisely the pattern reported by Harisson et al. (2000).

However, closer inspection of the data revealed certain differences between the spontaneous speech data and the probe data reported here. In the present study, the children with SLI used regular past-tense inflections with lower percentages than they did in spontaneous speech (see Oetting and Horohov 1997 for a similar finding for English). As seen in Table 2, the children with SLI used these inflections in only 37% of obligatory contexts on average, whereas their percentage in spontaneous speech averaged 86%. This suggests that when the children with SLI had less freedom to select the verbs to use in past tense, they experienced greater difficulty than indicated by the Hansson et al. (2000) data.

The percentages of regular past-tense inflection use for the ND-MLU children did not show such large differences between probe responses and spontaneous speech. As seen in Table 2, the mean percentage on the probe items was 87% whereas the corresponding percentage reported by Hansson et al. (2000) for spontaneous speech was 99%. However, a caution is in order. As noted earlier, a smaller number of ND-MLU children than children with SLI were willing to participate in the probe tasks. Therefore, it is possible that the ND-MLU children who did participate were not representative of the group as a whole. However, the fact that the pattern of differences and similarities between the SLI and ND-MLU groups was the same for the spontaneous speech samples (with N = 14 in each group) and the probe tasks (with N ranging from seven to ten for the ND-MLU group and thirteen to fourteen for the SLI group) suggests that the lower Ns for the ND-MLU group were probably not distorting the findings.

The errors committed in regular past-tense contexts also differed somewhat in spontaneous speech and the probe tasks. Hansson et al. (2000) found that 95% of the errors by the children with SLI were productions of infinitives in place of regular past-tense inflections. In contrast, in the present study, we found that only 42% of the errors were substitutions of the infinitive form. This was not even the most frequent error type, as 54% of the errors were substitutions of the present-tense inflection.

We find no reason to suspect that task effects were responsible for the pattern of regular past-tense use observed in this study. For example, although the children with SLI showed considerably lower percentages of use of regular past on our probe task than in spontaneous speech, the ND-MLU children's use of regular past tense did not show a comparable gap between the probes and spontaneous speech. Given the younger age of the ND-MLU children, these children--rather than the children with SLI--might have been expected to have greater difficulty handling a more structured task.

We also considered the possibility that the findings were an accidental result of our selection of probe items that were not comparable in difficulty across the morpheme types. For example, it might have been the case that the verbs used for the regular past-tense probe items were less familiar to the children with SLI than those used for present tense. However, this seems unlikely. Five of the verbs were used in both the present-tense and the regular past-tense probes. The children with SLI used these verbs in present tense with a mean percentage correct of 80.9%; in past tense, their mean percentage correct was only 42.6%. These values correspond closely to those shown in Table 2 (86.7% and 37.4%, respectively) for the entire list of verbs.

It also appears doubtful that the verbal information provided by the examiner during the probes could be the source of the errors committed by the children. For each past-tense item, the examiner provided a model of a past-tense form (of a different verb), as shown in (3). The cues provided with such a procedure could have had a facilitative effect, by either "reminding" the children to use past tense, or by serving as a morphological prime (e.g. Leonard et al. 2000). Yet the children often used a present-tense form in these contexts.

The finding of present-for-past-tense substitutions on the probes raises the possibility that the spontaneous speech samples overestimated both the regular past-tense inflection abilities and the present-tense inflection abilities of the children with SLI. Present-tense inflections can appear in the same sentence contexts as regular past-tense inflections, unless a temporal adverb referring to the past (e.g. `yesterday') is also used. Therefore, the identification of a context for regular past tense must depend heavily on the nonlinguistic context and/or the linguistic context outside of the child's own sentence. In cases where the surrounding context was not clearly one of past tense, the children might have been credited with appropriate present-tense inflection use. This would add to the number of "correct" productions of present-tense inflections and decrease the number of incorrect productions in regular past-tense contexts.

However, if this occurred, it probably did not disproportionately affect the scores of the children with SLI. In the present study, 69% of the errors in regular past-tense contexts by the ND-MLU children were productions of present-tense inflections. Thus, these children might have been even more prone to present-for-past substitutions than the children with SLI. If so, the ND-MLU group's percentages for both present and regular past-tense inflections in spontaneous speech were at least as likely to be inflated as were the corresponding percentages for the SLI group.

Although the children with SLI used regular past-tense inflections on the probes with lower percentages than the ND-MLU children, there was clear evidence that the children possessed some degree of knowledge of these forms. This conclusion is not limited to the four children with SLI who scored 100% on the regular past-tense probes. Most of the children with SLI produced overregularizations of past. Productions such as springde `runned' for sprang `ran' suggest that the children were capable of using inflections with verbs they had never heard inflected in that way before.

Task factors do not appear to be responsible for this finding. For several of the irregular-past probe items, the examiner's model employed an irregular past form rather than a regular past form. Yet overregularizations of past were nevertheless observed. Furthermore, an inspection of the same children's spontaneous speech in Hansson et al. (2000) revealed a total of twenty overregularizations of past from nine different children with SLI. A similar occurrence of overregularizations that did not seem related to degree of regular past-tense use is reported by Serratrice et al. (this issue) in a study of English-speaking children with SLI (younger than the ones in the present study).

We also found evidence for creative use of present-tense inflections in the present study. Three of the fifteen verbs in the present-tense inflection probe have a present-tense form that is identical to the stem. Yet, five children with SLI produced responses such as horer for hor `hears', in which a present-tense inflection -er was added unnecessarily. On the other hand, the children with SLI were relatively poor at applying present-tense inflections and, especially, regular past-tense inflections to nonce words.

How can we resolve the discrepancy between, on the one hand, clearly productive use of present and past-tense inflections through overregularizations, and, on the other hand, a relative weakness in applying these same inflections to nonce words? This question is especially vexing given that the children with SLI were not only productive in their use of present-tense inflections but also comparable to the ND-MLU children in their percentages correct for these inflections on the probes. There are two possible sources for this discrepancy. One possibility is that children with SLI require an unusually long period of time between the time they first acquire a verb and the time they show an ability to inflect it. A second possibility is that children with SLI require longer to acquire the verb itself.

Evidence supporting the first possibility can be seen in Table 1. If children with SLI require a longer period between their initial acquisition of a verb and the point when they inflect the verb, a higher percentage of their failures to inflect verbs should appear in the "never correct" category than is seen for the ND-MLU children. That is, a higher percentage in the "never correct" category would suggest that the children had acquired a larger proportion of verbs without yet having acquired an ability to inflect them. As can be seen in Table 1, for the children with SLI, four of the 28 verbs (14%) showing failures to apply the present-tense inflection were cases of "never correct," whereas for the ND-MLU children, not one of the nineteen verbs showing failures to apply the present-tense inflection was a case of "never correct." (Recall, however, that even for the children with SLI, the cases of "sometimes correct" were too numerous to assume that when inflections occurred they were the result of memorization.) Because the ND-MLU children were so accurate in their past-tense inflection use, this comparison could not be extended beyond present-tense inflections.

Evidence for the second source for the discrepancy--that children with SLI require an unusually long period of time to acquire the verbs themselves--can be found in studies on English. According to this work, children with SLI have smaller lexicons than same-age peers, with verbs showing the largest gap between these groups (e.g. Fletcher and Peters 1984; Watkins et al. 1993). Even compared to younger MLU control children, children with SLI use fewer verb types (Conti-Ramsden and Jones 1997) or verb tokens (Jones and Conti-Ramsden 1997). Studies in which children's acquisition of new words is measured after only a few exposures show that children with SLI have greater difficulty than peers in retaining verbs in particular (e.g. Rice et al. 1994).

These two possibilities are not mutually exclusive. In fact, Leonard et al. (1999) found that a group of children with SLI employed fewer verbs in their speech than expected for their age, but the children's use of finite verb morphemes such as present and past-tense inflections was even more limited than would be predicted by the size of their verb inventory. Such a finding is consistent with slow acquisition of verbs in the first place, followed by slow application of inflections to these verbs. A similar suggestion is made by Serratice et al. (this issue), pointing to the importance of vocabulary-learning skills for the acquisition of verb morphology.

One of the most important findings of the present study was the significant difference between the children with SLI and the ND-MLU children in the use of regular past-tense inflections but not present-tense inflections. Hansson et al. (2000) reported the same finding for spontaneous speech. However, because divergent results for present and regular past-tense inflections have not been reported for English or other Germanic languages, it was important to replicate this pattern using other types of data, as was done in the present study. It would appear that, in Swedish, children with SLI are more likely to lag behind MLU control children in the use of regular past inflections than in the use of present-tense inflections.

To our knowledge, this finding is not compatible with most of the accounts of grammatical deficits in SLI that have been offered to date. As noted at the outset of the paper, the prosodic characteristics and average durations of present and regular past inflections provide no basis for predicting a difference in the results for these two inflection types according to the surface account. According to this account, irregular past is the only verb-morpheme type that should be produced as proficiently by the children with SLI as by the ND-MLU children. It is certainly true that this account predicts that the first past-tense forms acquired by children with SLI will be irregular forms, owing to the tense information contained in the stressed vowels in these verbs. However, the surface account has no explanation for why present and regular past-tense inflections produced different results.

The agreement-deficit account provides no basis for predicting these results, as Swedish verb paradigms involve distinctions in tense only, not agreement. The extended optional-infinitive account is more applicable to Swedish in that it assumes that children with SLI may treat tense as well as agreement as optional. That is, tense and agreement may both be specified, one (tense or agreement) may be specified and the other underspecified, or both may be underspecified. If tense is specified, the child selects the form with the [+past] or the [-past] (= present) feature. If tense is underspecified, the child selects a nonfinite form such as an infinitive. The latter assumption constitutes an advantage of this account in dealing with our Swedish data. The children sometimes used infinitives in present and past-tense contexts, and such use can be viewed as an utterance in which tense was underspecified. However, according to this account, children with SLI presumably have knowledge of the [+past] and [-past] features of tense such that when tense is specified in the utterance, the selected morpheme should be correct in its details. Yet, present-tense inflections were even more likely to replace regular past-tense inflections than were infinitives.

Although in its current formulation, the extended optional-infinitive account has no provisions for handling present-for-past-tense errors, one particular assumption in this account may provide the basis for capturing such errors without violating the basic nature of the approach. In his detailed formulation of the extended optional-infinitive account, Schutze (1997) assumed that all child (and adult) utterances contain the functional projection TP, and that tense encodes (at least) the binary distinctions [+/-past] and [+/-finite]. The presence of [+/-finite] features separate from [+/-past] features is based on the assumption that some languages may possess, for example, a distinct infinitive form that is used for past events, requiring [+past, -finite]. For English, it is assumed that when neither [+past] or [-past] is specified, a nonfinite form will be used. However, given the fact that [+/-past] and [+/-finite] are separate features within tense, it could be the case that a feature value is specified for finite but past is underspecified.

Given these observations, one possibility is that Swedish-speaking children may use "infinitives" when the utterance is underspecified both for past and for finite. When the utterance is specified as [+past], the past-tense form will be used, and when the utterance is specified as [-past], the present-tense form will be used. As in English, specification of [+past] or [-past] implies [+finite]. However, the reverse might not be true. In Swedish child language, utterances might be underspecified for past but specified as [+finite]. If so, the present-tense inflected form might be selected as the default.

The above elaboration of the extended optional-infinitive account rides on the assumption that some of the present-tense inflections observed in the speech of the children with SLI reflected [+finite] only. However, until additional data supporting such as assumption is obtained, we must consider the more straightforward possibility that the children's use of present-tense inflections consistently reflected specification of [-past], whereas specification of [+past] was more sporadic. Although the precise reasons for such a present tense-past tense asymmetry are not yet clear, close inspection of the literature suggests that it could be quite real.

Although Swedish may be the first Germanic language studied in which children with SLI differ from MLU controls in regular past but not present-tense inflections, it is not the first Germanic language showing considerable use of present-tense inflections in past-tense contexts. In a recent study on Dutch, de Jong (1999) found that a group of children with SLI were more likely to produce present-tense forms in past-tense contexts than to produce infinitives. A group of younger normally developing children, in contrast, rarely committed errors of either type. Marchman et al. (1999) also found a higher incidence of present-tense responses in English-speaking children with SLI compared to age-matched controls in a probe task eliciting past-tense forms.

Outside the boundaries of Germanic languages, there is also evidence of differences between children with SLI and MLU controls in regular past-tense forms without a corresponding difference in the use of present-tense forms. Dromi et al. (1999) reported this pattern in a study on Hebrew. The discrepancy between the findings for present and past-tense forms was attributed to differences in the complexity of the present and past-tense paradigms of the language. In Hebrew, present tense makes distinctions for number and gender, whereas past tense makes distinctions for person as well as number and gender. Errors on past-tense forms were usually productions of forms that shared most of the features of the correct form. In many instances, the children were able to express correct tense; for example, past second person feminine singular forms were often replaced by past third person feminine singular forms. However, in other instances, the children maintained appropriate agreement features, but produced the present-tense equivalent (e.g. present feminine singular). Although the number of features required in the past-tense paradigm might have been the determining factor in the children's difficulty, it is also possible that the difficulty was attributable to a limitation in the children's ability to mark agreement while simultaneously expressing past tense.

In English, third person singular -s and regular past -ed are used with similar, low percentages of use by children with SLI. This finding might easily be interpreted to mean that English-speaking children with SLI have similar difficulty with both present and past tense, and that the same might be expected for children with SLI who are acquiring languages similar to English in typology. However, we believe that such an interpretation would be premature. First, even within the extended optional-infinitive framework, English-speaking children might produce an utterance such as Marion run fast all the time either because they fail to mark agreement or because they fail to mark tense (e.g. Wexler et al. 1998). Thus, it is not always the case that tense in particular is the source of the error. Second, it is questionable that third singular -s marks present tense in any straightforward way. This inflection is typically used to express aspectual notions such as habitual activity (e.g. Mom drives me to school) or other general conditions that hold, whether they are occurring at the time of the utterance or not (e.g. Dad talks fast). Whereas in many languages, simple present is used to describe activities in pictures unless the ongoing nature of the activity is being emphasized, in English, the present progressive, rather than the simple present is the default. It seems possible to us, then, that in languages in which the present-tense inflection differs from the past-tense inflection only in the present-past distinction and not also in its agreement and/or aspectual properties, the two will not be comparable in difficulty. We suspect that our Swedish data are illustrating this point.

Lund University

Purdue University

Table 1. Expected and observed frequency for verb types
appearing at least twice in obligatory contexts

 Verb-use categories
Inflection Always Never Sometimes
 correct correct correct

Children with SLI
 Present
 expected 94.18 7.65 11.05
 observed 85.00 4.00 24.00
 Regular past
 expected 15.35 2.87 1.87
 observed 14.00 1.00 5.00
ND-MLU children
 Present
 expected 86.24 6.14 10.14
 observed 82.00 0.00 19.00

Table 2. Percentages of correct use of verb forms on the probes

Verb form SLI ND-MLU ND-A

Copula
 M 76.77 (a) 95.50 (b) 99.43 (b)
 SD 24.99 7.23 2.14
 N 13 8 14
Present tense
 M 86.71 (a) 94.50 98.50 (b)
 SD 16.58 8.04 4.05
 N 14 8 14
Regular past tense
 M 37.36 (a) 87.40 (b) 98.93 (b)
 SD 45.28 16.81 2.73
 N 14 10 14
Irregular past tense
 M 45.23 (a) 66.14 (a) 98.08 (b
 SD 36.13 35.01 4.80
 N 13 7 13

Note: Group differences are indicated by differing superscript letters.

Table 3. Summary of the children's use of each inflection with a
nonce word

 SLI ND-MLU ND-A
 no. freq no. freq no. freq

Present 3/11 8 4/5 14 13/13 37
Regular past 2/7 2 6/6 13 12/13 21
Infinitive 6/6 11 5/5 10 14/14 42

Note: "no." refers to the number of children using the inflection out
of the number whose responses obligated the use of the inflection;
"freq" refers to the total frequency of items showing use of the
inflection.

Note

(1.) The research reported here was supported by Research Grant 5 R01 DC 00-458 from the National Institute on Deafness and Other Communication Disorders, National Institutes of Health, USA. We would like to thank Ulrika Nettelbladt, responsible for the Swedish part of the project, for important contributions throughout the planning and execution of the study. We also thank Eva-Kristina Salameh, Britt Hellquist, and Ulrika Guldstrand for their part in testing the children, and for transcribing parts of the recordings. Finally, we would like to thank the children who participated, and the speech-language clinicians who helped us identify children for the study. Correspondence address: Kristina Hansson, Department of Logopedics and Phoniatrics, Lund University Hospital, S-221 85 Lund, Sweden. E-mail: Kristina. Hansson@logopedi.lu.se.

References

Andersson, Karin; and Emenius, Kerstin (1970). Barns tillampning av sprakliga regler. Unpublished paper, Department of Linguistics, Lund University.

Bartke, Susan (1994). Dissociations in SLI children's inflectional morphology: new evidence from agreement inflections and noun plurals in German. Paper presented at the Meeting of the European Group for Child Language Disorders, Garderen, The Netherlands.

Borenstein, Michael; and Cohen, Jacob (1988). Statistical Power Analysis: A Computer Program. Hillsdale, NJ: Erlbaum.

Clahsen, Harald (1989). The grammatical characterization of developmental dysphasia. Linguistics 27, 897-920.

--; Bartke, Susan; and Gollner, Sandra (1997). Formal features in impaired grammars: a comparison of English and German SLI children. Journal of Neurolinguistics 10, 151-171.

--; and Hansen, Detlef (1997). The grammatical agreement deficit in specific language impairment: evidence from therapy experiments. In The Inheritance and Innateness of Grammars, Myrna Gopnik (ed.), 141-160. Oxford: Oxford University Press.

Cleave, Patricia; and Rice, Mabel (1997). An examination of BE in children with specific language impairment: the role of contractibility and grammatical form class. Journal of Speech, Language, and Hearing Research 40, 480-492.

Cohen, Jacob (1988). Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ: Erlbaum.

Conti-Ramsden, Gina; and Jones, Melanie (1997). Verb use in specific language impairment. Journal of Speech, Language, and Hearing Research 40, 1298-1313.

de Jong, Jan (1999). Specific language impairment in Dutch: inflectional morphology and argument structure. Groningen Dissertations in Linguistics 28.

Dromi, Esther; Leonard, Laurence; Adam, Galit; and Zadunaisky-Ehrlich, Sara (1999). Verb agreement morphology in Hebrew-speaking children with specific language impairment. Journal of Speech, Language, and Hearing Research 42, 1414-1431.

Fletcher, Paul; and Peters, Jo (1984). Characterizing language impairment in children: an exploratory study. Language Testing 1, 33-49.

Gopnik, Myrna; and Crago, Martha (1991). Familial aggregation of a developmental language disorder. Cognition 39, 1-50.

Hadley, Pamela; and Rice, Mabel (1996). Emergent uses of BE and DO: evidence from children with specific language impairment. Language Acquisition 5, 209-243.

Hamann, Cornelia; Penner, Zvi; and Lindner, Katrin (1998). German impaired grammar: the clause structure revisited. Language Acquisition 7, 193-245.

Hansson, Kristina; and Nettelbladt, Ulrika (1995). Grammatical characteristics of Swedish children with SLI. Journal of Speech and Hearing Research 38, 589-598.

--; Nettelbladt, Ulrika; and Leonard, Laurence (2000). Specific language impairment in Swedish: the status of verb morphology and word order. Journal of Speech, Language, and Hearing Research 43, 848-864.

Hellquist, Britt (1982). Sprakligt Impressivt Test for Barn. Loddekopinge, Sweden: Pedagogisk Design.

Holmberg, Eva; and Bergstrrm, Anneli (1996). ORIS. Loddekopinge, Sweden: Pedagogisk Design.

--; and Stenkvist, Harriette (1983). Nya Lundamaterialet: Kartlaggning och Bedomning av Barns Sprakliga Formaga. Malmo: Utbildningsproduktion.

Jones, Melanie; and Conti-Ramsden, Gina (1997). A comparison of verb use in children with SLI and their younger siblings. First Language 17, 165-193.

Leissner, Inga; Nilsson, Birgitta; Nystrom, Gisela; and Wastesson, Birgitta (1962). Leiters Klosstest. Stockholm: Psykologiforlaget.

Leonard, Laurence; Eyer, Julia; Bedore, Lisa; and Grela, Bernard (1997). Three accounts of the grammatical morpheme difficulties of English-speaking children with specific language impairment. Journal of Speech, Language, and Hearing Research 40, 741-753.

--; Miller, Carol; and Gerber, Erika (1999). Grammatical morphology and the lexicon in children with specific language impairment. Journal of Speech, Language, and Hearing Research 42, 678-689.

--; Miller, Carol; Grela, Bernard; Holland, Audrey; Gerber, Erika; and Pentucci, Marcia (2000). Production operations contribute to the grammatical morpheme limitations of children with specific language impairment. Journal of Memory and Language 43, 362-378.

Loeb, Diane Frome; and Leonard, Laurence (1991). Subject case marking and verb morphology in normally-developing and specifically-language-impaired children. Journal of Speech and Hearing Research 34, 340-346.

Marchman, Virginia; Wulfeck, Beverly; and Ellis Weismer, Susan (1999). Morphological productivity in children with normal language and SLI: a study of English past tense. Journal of Speech, Language, and Hearing Research 42, 206-219.

Marcus, Gary; Pinker, Steven; Ullman, Michael; Hollander, Michelle; Rosen, T. John; and Xu, Fei (1992). Overregularization in language acquisition. Monographs of the Society for Research in Child Development 57(4), serial no. 228.

Miller, Carol; and Leonard, Laurence (1998). Deficits in finite verb morphology: some assumptions in recent accounts of specific language impairment. Journal of Speech, Language, and Hearing Research 41, 701-707.

Miller, John; and Chapman, Robin (1986). SALT: a computer program for the systematic analysis of language transcripts. Unpublished manuscript, Language Analysis Laboratory, Waisman Center, University of Wisconsin-Madison.

Oetting, Janna; and Horohov, Janice (1997). Past-tense marking by children with and without specific language impairment. Journal of Speech, Language, and Hearing Research 41, 62-74.

Ratner, Nan Bernstein; and Menn, Lise (2000). In the beginning was the wug: forty years of language-elicitation studies. In Methods for Studying Language Production, Lise Menn and Nan Bernstein Ratner (eds.), 1-23. Mahwah, NJ: Erlbaum.

Rice, Mabel; Noll, Karen Ruff; and Grimm, Hannelore (1997). An extended optional infinitive stage in German-speaking children with specific language impairment. Language Acquisition 4, 255-295.

--; and Oetting, Janna (1993). Morphological deficits of children with SLI: evaluation of number marking and agreement. Journal of Speech and Hearing Research 36, 1249-1257.

--; Oetting, Janna; Marquis, Janet; Bode, John; and Pae, Soyeong (1994). Frequency of input effects on word comprehension of children with specific language impairment. Journal of Speech and Hearing Research 37, 106-122.

--; and Wexler, Kenneth (1996). Toward tense as a clinical marker of specific language impairment in English-speaking children. Journal of Speech, Language, and Hearing Research 39, 1239-1257.

Schutze, Carson (1997). INFL in child and adult language: agreement, case and licensing. Unpublished doctoral dissertation, MIT.

--; and Wexler, Kenneth (1996). Subject case licensing and root infinitives. In Proceedings of the 20th Annual Boston University Conference on Language Development, vol. 2, Andy Stringfellow, Dalia Cahana-Amitay, Elizabeth Hughes, and Andrea Zukowski (eds.), 670-681. Somerville, MA: Cascadilla.

Ullman, Michael; and Gopnik, Myrna (1994). Past tense production: regular, irregular and nonsense words. McGill Working Papers in Linguistics 10, 81-118.

Watkins, Ruth; Rice, Mabel; and Molz, Candace (1993). Verb use by language-impaired children and normally developing children. First Language 37, 133-143.

Wexler, Kenneth; Schutze, Carson; and Rice, Mabel (1998). Subject case in children with SLI and unaffected controls: evidence for the Agr/Tns omission model. Language Acquisition 7, 317-344.

Received 20 March 2001

Revised version received 7 November 2001