Reciprocal relations between syntax and tense/agreement morphology in children’s interpretation of input: A look at children with specific language impairment

Introduction

Since the publication of the classic studies on the role of input (e.g., Newport, Gleitman, & Gleitman, 1977; Ninio & Bruner, 1978; Snow, 1972), much has been learned about how adults’ language can shape not only the time course but also the details of children’s language development. Even now, with our already large knowledge base (see Hoff, 2006 and Rowe, 2012 for reviews), new insights continue to emerge. For example, in recent years we have seen new ways to conceptualize input frequency (e.g., Ambridge, Kidd, Rowland, & Theakston, 2015), and statistical structure (e.g., Hills, 2013) in child-directed speech. We have seen how mothers’ overlap in words with their children’s prior utterances is more predictive of language growth than the children’s overlap with their mothers’ prior utterances (Che, Brooks, Alarcon, & Yannaco, in press). The number of different verbs in a parent’s input appears to be more predictive of their children’s verb diversity than the total number of verbs used by the parent (Hsu, Hadley, & Rispoli, 2017). There is now more recognition of the degree to which input effects depend on child intake (Rowe, 2015). The contribution of genetics to input effects is also better understood. Along with a parent’s linguistic input affecting the child, there also appear to be child-to-parent effects and shared genetic effects on parent and child (Dale, Tosto, Hayiou-Thomas, & Plomin, 2015).

One focus of recent input research has been the role of linguistic input in children’s development of morphosyntax. Earlier research suggested that children’s morphosyntactic development might benefit from a simpler grammatical input. However, it appears that longer rather than shorter input utterances may lead to more advanced morphosyntactic development in children (e.g., Hoff-Ginsberg, 1998; Huttenlocher, Vasilyeva, Cymerman, & Levine, 2002). Yet this observation comes with some important qualifications. In her review of the literature, Hoff (2006) made some especially insightful observations in this regard:

All of the observed benefits of complexity were obtained within the range of complexity in child-directed speech. Furthermore, children may filter out, by not processing, input that is too complex – with no negative consequences to language development – so long as sufficient processable input is available. In contrast, children have no way to make up for input that is too simple. (p. 75)

The present article takes as its starting point the assumption that, within the range of typical child-directed speech, somewhat more complex input is, on balance, beneficial but there can be negative collateral effects. These can take the form of morphosyntactic errors in children’s speech that are cases of the children misinterpreting (or ‘filtering out’) certain details of more complex input.

Much of the research devoted to presumed input-based morphosyntactic errors has been concerned with the (inconsistent) use of tense and agreement morphology by children acquiring English and other Germanic languages. The evidence has come from studies employing computational models, corpus data, and experimental designs. In studies using computational models, the models were designed with a limited processing span (simulating the limited spans assumed in young children), and a bias towards material at the end of the input sentence, simulating a recency effect (Croker, Pine, & Gobet, 2000, 2001; Freudenthal, Pine, Aguado-Orea, & Gobet, 2007; Freudenthal, Pine, & Gobet, 2006, 2009, 2010). When adult, child-directed utterances in English were used as input to the model, the model’s output approximated young children’s productions, with heavy use of bare-stem verb forms. This occurred because nonfinite verbs in English (which have bare stems) are heavily distributed toward the ends of utterances. When the processing span of the model was gradually increased to simulate children’s development, a higher percentage of the verbs in the output showed tense/agreement marking. Further evidence for the role of input came when these investigators supplied the model with child-directed input in Dutch, French, German, or Spanish. The resulting output approximated the early use of tense/agreement inflections by children acquiring these respective languages, with the highest proportion of tense/agreement forms seen when Spanish input was used.

Some of the corpus-based studies concerned with input effects have examined the errors made in children’s sentence productions. For example, Kirjavainen, Theakston, and Lieven (2009) found that errors such as me-for-I (e.g., Me do it) were closely associated with adult input containing the use of me in preverbal position (e.g., Let me do it). Furthermore, the particular verbs used in these ungrammatical productions by the children (as in do in this example) were closely tied to the particular verbs most often used in the adults’ utterances containing preverbal me.

Räsänen, Ambridge, and Pine (2014) found a verb-specific input relationship of a different kind. These investigators observed that verbs most likely to appear (correctly) in bare-stem form in adult English were those more likely to be produced (incorrectly) as bare stems in third person singular –s contexts by young children.

Experimental studies have often used novel verbs in varying syntactic frames as input. Theakston, Lieven, and Tomasello (2003) presented children with novel verbs appearing strictly in a third person singular –s context, as in This one tams, exclusively in a question context in which the novel verb was nonfinite (because finiteness is carried by the fronted modal auxiliary verb), as in Will this one keef?, or in both types of contexts. Following the exposure period, the children’s productions showed a significant bias towards using the novel verb in the manner in which it was heard (nonfinite or with third person singular –s), even when the novel verb was tested in a context requiring the form that was not modeled during the exposure period. Similar findings were reported by Finneran and Leonard (2010).

Why inconsistent but otherwise correct use?

One question that has not been answered satisfactorily is why English-speaking children’s slow development of adult-like tense/agreement use takes the form of either using the correct tense/agreement morpheme, or using a nonfinite verb in its place. Incorrect substitutions of one tense/agreement morpheme for another occur, but are quite infrequent. Within the generative linguistic approach of Wexler, Rice, and their colleagues, tense and agreement are present in children’s grammars from an early point, but a biologically-based linguistic constraint prevents adult-like consistency. Until the constraint is overcome, tense/agreement is optional in sentence use (e.g., Rice & Wexler, 1996; Rice, Wexler, & Hershberger, 1998; Wexler, 1994, 1998). When tense/agreement is not represented in a sentence, a nonfinite verb form is used instead. That is, a production such as Morgan like ice cream is not regarded as a failed attempt to add the inflection –s to the verb, but rather as the selection of a verb form lacking tense/agreement.

If such an account is true, it still leaves open the question of why development from limited but correct use to mastery proceeds at such a slow pace. Could it be the case that some of the gradual pace is due to children’s difficulty understanding particular syntactic structures in the input that contain nonfinite verb forms? In a sense, this must be an implicit assumption when researchers tie children’s incorrect use to details in the input. However, there are aspects of children’s use that suggest that certain structures might be more problematic than others. Note that errors such I wants, you wants, and they wants are not common, suggesting that when children do hear tense/agreement forms in the input, they can evidently associate these forms with the proper tense, number, and agreement features. Second, children’s utterances such as Him open the present – that on first blush seem quite exotic – are on closer inspection suspiciously similar to nonfinite clauses in adult utterances such as Help him open the present and I saw him open the present. It seems more parsimonious to assume that such utterances are based directly on input rather than generated from scratch by the child. The latter requires two assumptions – that a nonfinite verb (e.g., open) was selected, and that the lack of tense/agreement in the verb forced the use of a default pronoun (e.g., him), rather than a pronoun of nominative case (e.g., he).

In other languages, a ‘generated-from-scratch’ view must make similar additional assumptions. For example, in German, children’s use of a nonfinite verb in place of a verb marked for tense/agreement often results in the verb appearing in sentence-final position, as in Das Mädchen das Geschenk öffnen (‘The girl the gift open’) instead of in second position, as in Das Mädchen öffnet das Geschenk (‘The girl opens the gift’). A generated-from-scratch account would hold that, although the children were inconsistent in using tense/agreement, they nevertheless ‘knew’ that nonfinite verbs appear at the end of the sentence. Yet, in German, an input utterance such as Kann das Mädchen das Geschenk öffnen? (literally, ‘Can the girl the gift open?’ = ‘Can the girl open the gift?’) contains the sequence das Mädchen das Geschenk öffnen.

In this review, we describe some of the syntactic structures that, if misunderstood, might mislead children into extracting portions of the input and using them as free-standing utterances and as the basis for generating new utterances with the same errant structure. In doing so, we adopt many of the assumptions of the ‘competing-sources-of-input’ approach (see Leonard, 2014a, pp. 331–341) – an approach aimed at accounting for the optional use of tense/agreement morphemes that is so well documented in the child language literature. Here, we will emphasize the optional tense/agreement morpheme use of a group of children who are especially slow in reaching mastery in the use of these morphemes – children with specific language impairment – and the role that input might play in this extended inconsistent use.

Children with specific language impairment

Children with specific language impairment (SLI) are children with significant deficits in language ability that cannot be attributed to any known cause. Their hearing is within normal limits, their scores on tests of nonverbal intelligence are safely above the intellectual disability level, they show no clear evidence of neurological damage, and they do not meet the criteria for autism spectrum disorder. In most cases, there appears to be a genetic contribution to the language impairment, though it is also clear that this disorder is multifactorial. Although the term ‘specific language impairment’ is by far the most commonly used label for these children in the research literature (Bishop, 2014a; Leonard, 2014b), it has the drawback of implying a ‘pure’ language deficit. In fact, although language is clearly the major deficit in SLI, subtle weaknesses in other developmental areas can also be seen in many cases (see Leonard, 2014a for a review). ‘Developmental language disorder’ is an alternative clinical label that has recently been applied to these children (Bishop, Snowling, Thompson, Greenhalgh, & CATALISE-2 Consortium, in press).

Because language seems to be the central area of these children’s difficulties, early studies asked whether their linguistic input was in some way insufficient to allow for typical language development. However, with the help of some sophisticated research designs, investigators demonstrated that the input was unlikely to be the major contributor to these children’s problems (e.g., Conti-Ramsden & Dykins, 1991; Cross, Nienhuys, & Kirkman, 1985). In these studies, children with SLI were matched with younger typically developing children according to mean length of utterance (MLU). Furthermore, some of the younger typically developing children were the younger siblings of the children with SLI. Thus, not only could parents of children with SLI be compared to parents of typically developing children showing MLUs similar to that of the children with SLI, but the input of the same parent could also be examined when interacting with her child with SLI and when interacting with the MLU-matched younger sibling. The results of these studies revealed far more similarities than differences in the input directed at children with SLI and younger typically developing children. These findings make it unlikely that a flawed input is driving the language impairment in children with SLI. On the other hand, given the genetic connections involved in SLI, it is very plausible that subtle effects are at work in parent–child interactions, including shared genetic effects operating on both parent and child (Dale et al., 2015).

Rather than the input serving as the cause of the impairment, it is the children’s intake of the input that seems the more likely source of the problem (see Rowe, 2015). That is, although misinterpretation of complex input might lead to short-lived morphosyntactic errors in typically developing children, such errors will persist for longer periods in children with SLI because these children will be very slow to resolve their comprehension difficulty with this type of input.

If misinterpretation of the input is one source of the morphosyntactic difficulties in children with SLI, then grammatical morphology in general and tense/agreement morphology in particular are likely to be the main casualties. Comparative studies reveal that preschool-age children with SLI are weaker in using tense/agreement morphology than are typically developing children two years younger, even when the groups are matched for MLU or the size of their verb inventory (Conti-Ramsden, Botting, & Faragher, 2001; Leonard, Eyer, Bedore, & Grela, 1997; Leonard, Miller, & Gerber, 1999; Oetting & Horohov, 1997; Rice & Blossom, 2012; Rice, Wexler, & Cleave, 1995). Although tense/agreement morphology has been the most widely studied aspect of these children’s grammatical morpheme difficulties, children with SLI have also been shown to be weaker than younger typically developing children in their use of possessive ’s, nonthematic of, the articles a and the, infinitival to, and the obligatory relativizer that, among others (Leonard, 1995; Owen & Leonard, 2006; Polite, Leonard, & Roberts, 2011; Schuele & Tolbert, 2001).

Competing sources of input and SLI

In a series of studies, Leonard and his colleagues have adopted an input-based approach – the competing-sources-of-input account – in an attempt to account for the extended optional use of tense/agreement morphemes in children with SLI (Leonard & Deevy, 2011; Leonard, Fey, Deevy, & Bredin-Oja, 2015; Purdy, Leonard, Weber-Fox, & Kaganovich, 2014; Souto, Leonard, Deevy, Fey, & Bredin-Oja, 2016). They assume that this protracted inconsistency is due to the combined effects of these children’s more general syntactic comprehension difficulties (Bishop, 2014b), and their reliance on communicatively functional subject–verb sequences that occur as parts of larger structures in the input. The premise is that many instances in which children fail to include a tense/agreement morpheme can be a natural consequence of misinterpreting particular types of utterances in the input. Thus, rather than assuming a problem with tense/agreement that is separate from these children’s well-documented limitations in other areas of language, this approach tries to integrate tense/agreement problems with the more general language problem.

The specific kinds of structures assumed to be related to tense/agreement inconsistency are those in which a matrix verb dictates the use of a nonfinite lexical verb later in the utterance, or a fronted auxiliary verb in a question is separated from the lexical verb that appears later. Examples include Help Mom do the dishes, We saw Dad wash the car, Make the boy take out the trash, Does the girl run fast?, and Is the dog eating?, among others. These are quite varied but have in common the presence of a subject–nonfinite verb sequence, that, on semantic grounds is communicatively functional (e.g., Mom do the dishes, Dad wash the car, The boy take out the trash, The girl run fast?, The dog eating?). The claim is not that such input utterances are incomprehensible to the children; rather, it is assumed that the children fail to understand the structural ties between the matrix verb/auxiliary verb and the later-appearing lexical verb. As a result, there is no structural constraint against extracting the subject–nonfinite verb sequence and using it both as a stand-alone utterance (e.g., Dad wash the car) and as a basis for generating new utterances (e.g., Mom wash the dog).

Simple declarative sentences are assumed to present fewer challenges of this type and serve as the source of the children’s use of utterances that include tense/agreement morphemes. Thus, utterances from the same child such as Timmy ride the bus and Molly rides the train could have different sources (as in Does Timmy ride the bus? and Every day, Molly rides the train). This assumption is what gives the approach the name ‘competing sources of input.’

Such competing utterances are part of a single grammar. For example, the utterances Her playing the piano and She is playing the piano might have their origins in input utterances such as We hear her playing the piano and She is playing the piano right now. However, because these alternatives are from the same grammar and generalizations can occur in each direction, even an utterance such as Her is playing is compatible with the children’s manner of speaking (Leonard, 2014a, p. 339). Although accusative pronouns used incorrectly as sentence subjects usually appear without a tense/agreement morpheme (as in Her playing), examples such as Her is playing occur more often than an assumption of ‘accidental’ error would predict (Pine, Rowland, Lieven, & Theakston, 2005).

Another assumption of the competing-sources-of-input approach is that the input structures housing subject–nonfinite verb sequences are sufficiently diverse to prevent a wholesale change in tense/agreement use once children finally learn to correctly interpret any one of them. Instead, as children learn the structural ties involved in each of these input types, there will be an incremental increase in the children’s use of tense/agreement morphemes. Viewed from the opposite perspective, if children show a high degree of tense/agreement morpheme use, they will prove to be less influenced by input structures containing subject–nonfinite verb sequences, because, presumably, they will have already learned the structural ties involved and no longer view the nonfinite sequences as appropriate to use on a stand-alone basis.

This last point is especially relevant for comparisons between children with SLI and younger typically developing children. For example, five-year-old children with SLI use tense/agreement morphemes with lower percentages of use (approximately 20–60% on average) than do three-year-old typically developing peers (approximately 55–85% on average). This difference leads to the prediction that although the younger typically developing children have not yet reached mastery, they will be less influenced by input than will children with SLI. That is, by virtue of their greater (albeit below-mastery) use of these morphemes, young typically developing children are probably on their way to properly understanding some of the structures in the input that house subject–nonfinite verb sequences and, as a result, no longer use them to generate utterances of their own. More of these structures will continue to be problematic for children with SLI and, for this reason, these children will remain at lower levels of tense/agreement use for a protracted period.

Syntax as a source of grammatical morpheme difficulties in SLI

This issue was first explored by Leonard and his colleagues through an adaptation of a task first employed by Theakston et al. (2003) with young typically developing children. In studies by Leonard and Deevy (2011) and Leonard et al. (2015), children were presented with novel verbs in sentence frames containing a tense/agreement morpheme and other novel verbs in frames with the novel verb in nonfinite form. (Novel verbs were, of course, counterbalanced so they appeared in each form across children.) Examples of contrasting frames were Just now the bird was channing vs. We saw the dog pagging; All day long the dog kreets vs. Let’s watch the bird tome; and Do you think the dog swopes? vs. Does the bird neff? Each verb was presented in one form only (with a tense/agreement morpheme or in nonfinite form). Following the exposure period, the children were tested for their use of each novel verb in a frame compatible with its presentation form, and in a frame requiring the children to use the novel verb in a different form. For example, after exposure to Let’s watch the bird tome, the children were tested with sentence completion items of the type Every day the cat (tomes) and We wanna watch the monkey (tome). Note that during testing, the children’s ability to use the novel verb with a different sentence subject was also assessed.

The results of these studies were quite clear. Five-year-old children with SLI whose tense/agreement morpheme use averaged around 35% were highly influenced by the manner in which the novel verb was heard during the exposure period. Three-year-old typically developing children whose tense/agreement use averaged around 60% were also influenced by the input, but less so than the SLI group. Typically developing children five years of age who were at mastery in their use of tense/agreement morphemes showed minimal input effects. Of note was the finding that the children with SLI produced errant forms in both directions. Upon hearing frames such as Does the bird neff?, they were prone to respond incorrectly to test items by saying Every day the horse (neff). In contrast, after hearing frames such as Do you think the dog swopes?, they often produced responses during testing such as We wanna watch the frog (swopes).

The novel verb studies showed especially strong input effects on the tense/agreement use of children with SLI but they were not a direct test of the assumption underlying this work – that these children do not adequately understand the syntactic structures housing subject–nonfinite verb sequences. One of the experiments reported in Leonard and Deevy (2011) and a more definitive study conducted by Souto et al. (2016) tested this assumption more directly. Across the two studies, four- and five-year-old children with SLI were compared with three- and four-year-old younger typically developing children who were matched on their scores on a general sentence comprehension test. The SLI group was inconsistent in their use of auxiliary is whereas the younger peers were near or at mastery in using this morpheme. The children were tested on their comprehension of sentences of the type The Dad sees the girl sleeping. This structure was assumed to be more problematic for the children with SLI, given that it contains a subject–nonfinite verb sequence (e.g., the girl sleeping) and these children’s inconsistent use of auxiliary is left open the possibility that structures of this type were contributing to the children’s continued treatment of auxiliary is as optional. To ensure that the children could understand the component parts of this structure, the children were also tested for their comprehension of sentences such as The girl is sleeping and The Dad sees the girl. A picture-pointing task was employed. In both studies, the children with SLI were less accurate than the younger typically developing children on the items containing subject–nonfinite verb sequences. The two groups were similarly accurate in responding to the component-parts control items.

Earlier, it was noted that in the novel verb studies, children with SLI were more likely than their peers to complete the experimenter’s sentence with a novel verb in third person singular –s form in contexts where a nonfinite form was expected, as in We wanna watch the frog (swopes). Such sentences do not appear in children’s input. On the other hand, a response such as swopes shows local agreement with the subject (the frog swopes) if one ignores the fact that a perception verb such as watch requires a nonfinite verb in the subordinate clause. To explore this issue further, Purdy et al. (2014) conducted a study involving both behavioral and event-related potential (ERP) measures with older children with SLI and same-age typically developing peers. The children with SLI were no longer producing errors in simple-sentence contexts, though occasional errors occurred in their use of more complex structures.

Purdy et al. (2014) asked whether the children with SLI would be less likely than peers to respond to errors such as that in He makes the boy talks a little louder. To detect the error, one must ignore the local agreement (the boy talks) and understand that a nonfinite verb is required in the subordinate clause. Fully grammatical control sentences such as He makes the boy talk a little louder were also included. In addition, Purdy et al. also included errors in simple sentences in which third person singular –s appeared inappropriately, as in Every night they talks on the phone. Along with examining the children’s behavioral (correct-incorrect) judgments of each sentence, Purdy et al. determined whether the errors elicited a characteristic P600 – an electrophysiological response interpreted as reflecting the detection of an error of a grammatical nature. For simple sentences of the type Every night they talks on the phone, the children with SLI did not differ from the typically developing children. However, on both the behavioral and electrophysiological measures, the children with SLI showed less sensitivity than their peers on sentences of the type He makes the quiet boy talks a little louder. The results were interpreted to indicate that the children with SLI were less secure in their understanding of the larger structure, and were therefore unduly influenced by the local agreement between the subject of the subordinate clause (e.g., the quiet boy) and the verb that followed (e.g., talks).

Collectively, these studies implicate poor comprehension of select syntactic structures in the input as a source for the protracted inconsistent use of tense/agreement morphology by children with SLI. As noted earlier, it seems possible that the very gradual increase seen in these children’s tense/agreement use in obligatory contexts is due in part to a structure-by-structure learning of the dependency relationships between the matrix verb/auxiliary verb and the nonfinite lexical verb appearing later in the utterance. Some of these structures might be learned before others, and each resolution might lead to an incremental change in the children’s tense/agreement use.

If it is determined that the dependency relationships in some structures are learned by typically developing children earlier than others, the former might be the first type of structure used in treatment studies with the SLI population. It is also likely that some structures exert a greater influence on children’s protracted tense/agreement inconsistency than others. One candidate might be questions with fronted auxiliaries, as in Can the dog catch the ball?, Did the cat go outside?, and Is the boy riding his bike? Unlike some of the other candidate structures, questions of this type are single-clause utterances, but they nevertheless require an understanding that the nonfinite lexical verb is only physically but not structurally separate from the fronted tense/agreement-marked auxiliary. Questions are frequent in the input of young children, representing over 30% of child-directed utterances on average (Cameron-Faulkner, Lieven, & Tomasello, 2003). Furthermore, they represent a very plausible source of children’s productions that combine a subject pronoun in nominative case with a nonfinite verb, as in She like apples and He need a spoon. Note that even with the intonation of a question, there is nothing in the extracted subject–nonfinite verb sequence that prevents it from being used in a declarative context. In fact, children hear otherwise identical utterances with different intonation when they hear casual questions (e.g., Mommy’s taking a nap?) that match the structure of declarative sentences.

Grammatical morphemes as cues to syntactic interpretation in SLI

One advantage of the competing-sources-of-input approach is that it integrates tense/agreement inconsistency with the broader problem of morphosyntax exhibited by these children. Part of this broader problem is a special difficulty with grammatical morphemes. As noted earlier, these morphemes include not only tense/agreement morphemes but also several morphemes that are not directly related to tense and agreement.

If grammatical morphemes as a larger class represent a problem area, it is likely that some portion of these children’s tense/agreement difficulty might be related to the morphemes themselves in addition to the competing contexts in which they appear or fail to appear. This could represent a vicious circle: Tense/agreement morpheme inconsistency may seem allowable to children because the children do not adequately grasp certain syntactic structures, and the children’s difficulty with the morphemes themselves may significantly delay the children’s ability to use these morphemes as aids in correctly interpreting the syntax.

There is no easy way out of this vicious circle if the focus is limited to tense/agreement morphemes. For example, additional experience with these morphemes in simple sentences will not inform the children that extraction and use of subject–nonfinite verb sequences from the input is ungrammatical. Yet presenting these morphemes in the challenging contexts presents its own complications. An example is seen in the study of Deevy, Leonard, and Marchman (2014, 2017). These investigators proposed that children with SLI would be more prone to extracting nonfinite subject–verb sequences from auxiliary-fronted questions if they did not understand the structural relation between the fronted auxiliary and the lexical verb. For example, if the children treated is in Is the baby crying? as simply serving the pragmatic function of signaling a question, there would be no constraint against extracting the baby crying and using it as a basis for new utterances. If, on the other hand, the structural ties between the fronted auxiliary and the lexical verb were understood, extraction would be less likely.

As a first step in exploring this idea, Deevy et al. (2014, 2017) used a looking-while-listening procedure. For each item, two pictures were presented on a large screen, such as a drawing of one boy running and another of two dogs running. While looking at the drawings, the children heard questions such as Is the nice little boy running? or Are the nice little dogs running? In contrasts such as these, the singular/plural information expressed in the fronted auxiliary allowed the children to shift their gaze towards the named picture before hearing the noun. The children’s eye gaze towards the appropriate picture for these contrasts was compared to their looking when only the noun provided the distinguishing information, as in See the nice little boy running? or See the nice little dogs running? Deevy et al. found that typically developing three-year-olds nearing mastery in their use of auxiliary is and are showed anticipatory looking based on the fronted auxiliary. This pattern of responding suggested that the children treated auxiliaries as an integrated component of the interrogative structure. In contrast, no anticipatory looking was noted in five-year-old children with SLI. These children did not appear to be treating fronted auxiliaries as components tightly tied to structure. Instead, these children appeared to treat the auxiliary as simply marking a question, not unlike the function served by question particles placed at the periphery of a sentence in certain Asian languages.

This type of problem seen in the children with SLI cannot be cleanly divided into a syntax problem and a tense/agreement morpheme problem. The children did not understand the structural ties, but this was exacerbated by their limited understanding of the full grammatical role played by auxiliaries.

Fey, Leonard, Bredin-Oja, and Deevy (2017) found one possible way of approaching this dual challenge, in an intervention study with a group of children with SLI. These investigators taught children in the experimental group to respond to questions such as Is the boy climbing the ladder? and Was the girl painting a fence? by attending to the tense information contained in the fronted auxiliary. This was accomplished by using alternative video sequences that differed only in whether the action was being performed or had just been performed. The children in the other group received exposure to the same questions, but the alternative video sequences differed in the actor performing the action (e.g., a girl rather than a boy climbing the ladder) and therefore provided no information about the grammatical role of the fronted auxiliary. Fey et al. found that the children in the experimental group made greater gains in their grammatical understanding of auxiliary-fronted questions, and also showed greater subsequent use of auxiliaries in their own speech. The latter finding was attributed to the children’s growing understanding (thanks to the intervention) that nonfinite subject–verb sequences are structurally connected to the fronted auxiliary.

There are other grammatical morphemes that might be helpful in serving as cues to the differences between structures that require the presence versus absence of tense/agreement morphemes. These include pronouns and the complementizer that. Each can be somewhat difficult for children with SLI, but, because they are not themselves tense/agreement morphemes, they might be put to use in helping children resolve their problem with particular syntactic structures. For example, for certain structures, the pronoun serving as a subject in the subordinate clause provides an additional cue. Compare the pair We hear the girl crying and We think the girl is crying with the pair We hear her crying and We think she is crying. Children might eventually learn the distinction in the first pair by recognizing (or being taught) that perception verbs such as hear require a nonfinite verb in the subordinate clause whereas cognition verbs such as think require a finite verb. This distinction might be enhanced in the second pair. That is, if children note not only the pairing between hear and crying but also between hear and the more proximal accusative pronoun her, learning might occur more readily. Similarly, the pairing between think and the proximal nominative pronoun she could facilitate learning in conjunction with the more distal think–is crying pairing.

The learning of other syntactic structures might also be assisted by the presence of accusative pronouns. It seems possible that pairings between matrix verbs such as help, make, and let and the (nonfinite) verbs that follow become stronger with intervening accusative pronouns, as in Help her do the dishes, Jason made me drop the ball, and Let him take a turn.

Another cue to whether a finite or nonfinite verb is used in the subordinate clause is the complementizer that. In particular contexts, a sentence such as We see the man is fixing dinner is perfectly grammatical. The meaning conveyed by this utterance is not identical to the meaning of We see the man fixing dinner, but each could be spoken in a very similar context, and therefore might be mistakenly assumed to be appropriate alternatives to express the same intent. The inclusion of the complementizer that helps to separate the two meanings, as in We see that the man is fixing dinner versus We see the man fixing dinner. Because this complementizer never permits a nonfinite clause to follow, with its inclusion, children might be able to distinguish the two structures at an earlier point in development.

Conclusions

Children with SLI and typically developing children differ in their genetic endowment for language. But, as noted by Dale et al. (2015, p. 116), ‘genes need the environment to have their effect.’ For typically developing children, a stronger genetic endowment might incline children to spend more time engaged in language with those around them, whereas, for children with SLI, such activity might be a slog, with particularly challenging aspects of the language input especially problematic for them to decipher. The challenging aspects of the input in English probably include the ubiquitous appearance of bare stems and the presence of nonfinite verbs that are separated in the sentence from the preceding matrix verbs or fronted auxiliaries that dictate their nonfinite form. Typically developing children learning this language work through these obstacles, causing only a moderate delay in their morphosyntactic development relative to children learning other languages. For children with SLI, on the other hand, breaking through these obstacles can take much longer. These intake problems translate into an extended period of tense/agreement inconsistency. An unfortunate circularity can also be created as a result of the children’s difficulty with problematic structures in the input: By assuming that tense/agreement and related morphemes are merely optional, children with SLI might fail to note the important cues to correct sentence interpretation that particular grammatical morphemes provide. Whether through enhanced input or formal intervention techniques, an emphasis on these morpheme cues might prove especially beneficial for these children.