Professional Documents
Culture Documents
Children's Responses To Telegraphic and Grammatically Complete Prompts To Imitate - Bredin-Oja and Fey 2014
Children's Responses To Telegraphic and Grammatically Complete Prompts To Imitate - Bredin-Oja and Fey 2014
Research Article
Purpose: The purpose of this study was to determine Results: No differences between conditions were found for
whether children in the early stage of combining words are the number of responses that contained a semantic relation.
more likely to respond to imitation prompts that are telegraphic In contrast, 3 of the 5 children produced significantly more
than to prompts that are grammatically complete and whether grammatical morphemes when presented with grammatically
they produce obligatory grammatical morphemes more reliably complete imitation prompts. Two children did not include a
in response to grammatically complete imitation prompts than function word in either condition.
to telegraphic prompts. Conclusion: Providing a telegraphic prompt to imitate does
Method: Five children between 30 and 51 months of age not offer any advantage as an intervention technique. Children
with language delay participated in a single-case alternating are just as likely to respond to a grammatically complete
treatment design with 14 sessions split between a grammatical imitation prompt. Further, including function words encourages
and a telegraphic condition. Alternating orders of the 14 sessions children who are developmentally ready to imitate them.
were randomly assigned to each child. Children were given
15 prompts to imitate a semantic relation that was either Key Words: children, early intervention, language disorders,
grammatically complete or telegraphic. morphology, syntax
M
any speech-language pathologists (SLPs) working To many, the use of telegraphic models in language
with families of young children with language intervention is intuitively sensible. Adults use telegraphic
delay shorten their utterances and encourage models based on the belief that the reduced input will lighten
parents and other professionals to shorten their utterances so the child’s comprehension burden by removing potentially
that they more closely match the child’s expressive language competing elements that perform principally formal gram-
level. When such shortening involves stripping grammatical matical functions, such as articles, auxiliaries, and bound
morphemes from otherwise acceptable phrases or sentences, morphemes. Such short utterances without grammatical detail
the result is reduced or telegraphic models. The following are presumed to make the meaning of the adult utterance
examples contrast a telegraphic utterance (a) followed by a more accessible, thus rendering them easier to imitate and,
corresponding grammatically complete utterance (b). ultimately, to produce spontaneously and creatively. On the
1a. boy in house basis of this logic, many language intervention approaches
have recommended the use of telegraphic input with children
1b. the boy is in the house with language impairments (e.g., Hancock & Kaiser, 2006;
2a. girl walk MacDonald & Blott, 1974). For these models of intervention,
2b. the girl is walking adult prompts for verbal imitation represent a key ingredient
and are assumed to be central to program success. Therefore,
3a. baby like milk a crucial assumption in these approaches that recommend the
3b. the baby likes milk use of telegraphic imitative stimuli is that utterances contain-
ing language intervention targets are more reliably imitated
when they are shortened by stripping them of grammatical
a
University of Kansas Medical Center, Kansas City
functors (i.e., function words) than they would be if they were
presented with all grammatical details intact. In this study,
Correspondence to Shelley L. Bredin-Oja: sbredin-oja@kumc.edu
we put this assumption directly to test. The purpose of this
Editor: Carol Scheffner Hammer
study, therefore, is not to test the impact of an intervention
Received December 7, 2012
Revision received May 7, 2013
Accepted June 15, 2013 Disclosure: The authors have declared that no competing interests existed at the
DOI: 10.1044/1058-0360(2013/12-0155) time of publication.
American Journal of Speech-Language Pathology • Vol. 23 • 15–26 • February 2014 • A American Speech-Language-Hearing Association 15
over time but, rather, to test the immediate response of a child were presented as recasts in interrogative form (e.g., Child:
when presented with a prompt to imitate. “Baby eat.” Adult: “Will the baby eat?”). The results indi-
cated that not only did the use of auxiliary-fronted questions
fail to facilitate auxiliary development but also the children in
Reasons to Question the Use of Telegraphic Models the experimental group used (numerically, but not statisti-
There are many reasons to believe that the use of tele- cally) fewer auxiliaries than did the children in the play group.
graphic speech may not be beneficial to young language On the basis of these results, the researchers reasoned that if
learners and that, in fact, telegraphic input may complicate children early in development do not fully process the aux-
learning processes. First, there is substantial evidence that iliary immediately before the sentence subject, frequent
children do not comprehend adult utterances more readily adult use of subject–auxiliary reversals may wrongly convey
when grammatical functors are removed (see Fey, 2008, and to the child that unmarked strings such as “the baby eat”
van Kleeck et al., 2010, for a review). With few exceptions are acceptable in main clauses. Theakston and colleagues
(e.g., Smith, 1973), these studies either indicate no signifi- (Theakston & Lieven, 2008; Theakston, Lieven, & Tomasello,
cant difference in comprehension between grammatical and 2003) provided experimental support for the hypothesis that
telegraphic sentences or favor the grammatical models unmarked strings found in auxiliary-fronted questions neg-
(van Kleeck et al., 2010). atively impact typically developing children’s use of finite
Second, it is well documented that children have the markers. Other researchers (e.g., Hadley, Rispoli, Fitzgerald,
most difficulty learning grammatical forms that are sparsely & Bahnsen, 2011; Leonard & Deevy, 2011) who considered
represented and/or optional in the language (cf. Leonard other forms in adult language have also concluded that input
et al., 2003; Leonard, Eyer, Bedore, & Grela, 1997; Rice & devoid of overt tense and agreement marking leads children
Wexler, 1996). For example, in English, use of tense and to consider forms that incorrectly lack tense and agreement
agreement inflections on verb forms has been shown to be as acceptable. If strings of words not fully processed within
difficult for English-speaking children with typical develop- grammatical input can lead the child to an incorrect hypoth-
ment and exceedingly difficult for children with specific esis about the obligatory nature of morphological markers,
language impairment (SLI; Leonard, Camarata, Brown, & what impact does telegraphic input have, especially in the case
Camarata, 2004; Leonard, Camarata, Pawłowska, Brown, & of imitative prompts in which the child is explicitly instructed
Camarata, 2006; Rispoli, Hadley, & Holt, 2012). One factor to produce an ungrammatical utterance? Finally, there
responsible for this delay appears to be that verbs are not is a great deal of evidence that children are aware of gram-
marked consistently for tense and agreement in English. For matical morphemes and recognize some of these morphemes’
example, past tense is unmarked on some forms (e.g., hit, put), grammatical functions well before they produce them (Eyer
lexically marked in others (e.g., went, hid), and inflected et al., 2002; Höhle & Weissenborn, 2003; Höhle, Weissenborn,
with a phonologically conditioned suffix in others (e.g., walked, Kiefer, Schulz, & Schmitz; 2004; Olguin & Tomasello, 1993;
posed). The third person singular –s is another obvious case Shi, Marquis, & Gauthier; 2006; Tomasello & Olguin, 1993).
in point; the marking of present tense and number is limited As one example, Fernald and Hurtado (2006) directly com-
to the third person. In contrast, despite the richness and pared 18-month-old toddlers’ recognition of words in contexts
complexity of verb morphology in many languages, such as with grammatical detail (e.g., look at the baby) to contexts
Italian and Hebrew, children with SLI who are learning these with all grammatical detail omitted (Look. baby.). The results
languages do not show a profile of special difficulty with in- showed that the children were significantly faster to respond
flected verb forms (Bedore & Leonard, 2001; Dromi, Leonard, accurately to target words when those words are presented
& Shteiman, 1993; Leonard & Bortolini, 1998). Use of tele- in a short grammatical sentence. Thus, it appears that com-
graphic models has the effect of making the overall use of plete phrases and sentences with preserved prosodic contours
grammatical forms less consistent. Rather than simplifying and predictable co-occurrences of forms, such as determiners
language learning, use of telegraphic adult input may provide with nouns, offer an advantage rather than a disadvantage
positive evidence to children that grammatical details are to young children learning language.
optional rather than obligatory, thus making them more
difficult to learn.
Is Telegraphic Speech Necessary for Improving
Third, there is evidence that even certain types of
grammatically correct input can have negative consequences Expressive Language?
if they are not fully processed by young language learners. Given the evidence that telegraphic input to young
In their study of the role that yes–no questions have on the children may be contraindicated and may even slow down the
acquisition of auxiliaries, Fey and Loeb (2002) hypothesized rate of language development in some cases, then what are the
that exposing children to sentence-initial auxiliaries would effects of regularly omitting them in language intervention?
increase the saliency and, therefore, lead to greater auxiliary There is no question that language interventions that require
development. The researchers exposed 3-year-old children children with language impairment to repeat telegraphic
with SLI and 2-year-old typically developing children who stimuli can be efficacious. For example, enhanced milieu
were not yet producing auxiliaries to either play sessions in teaching (Hancock & Kaiser, 2006), which advocates the use
which the adult produced frequent auxiliaries in declarative of telegraphic prompts for imitation and explicitly teaches
sentences or an experimental protocol in which auxiliaries parents to use these forms, has been widely studied and shown
Participant ID Sex Race/Ethnicity Age (in mos) MLUm/100 utterances Brown’s stage Leiter–R Brief IQ CDI:WS
Note. mos = months; MLU/m = mean length of utterance in morphemes; CDI:WS = MacArthur–Bates Communicative Development Inventories:
Words and Sentences; vocabulary checklist, total words used.
2010). For an ATD with two levels, the number of ways The grammatically complete version of this example (i.e.,
in which a unique random order occurs is calculated by the Say, the duck is walking) was never used in the grammatical
following formula: condition. Similarly, a child who received the stimulus,
Say, the boy is jumping, in the grammatical condition would
Number of ways ¼ N!=n1 !n2 !; never encounter the model, Say, boy jumping, in the tele-
graphic condidition. To facilitate this separation of tokens
where N = total number of sessions, n1 is the number of of the child’s target across conditions, different sets of toys
sessions for Treatment A, and n2 is the number of sessions were used for each condition.
for Treatment B. The children were seen in their homes or at their day-
Fourteen sessions of two equal numbers of prompt care either two or three times per week, with at least 1 day
conditions results in 3,432 possible random orders [14!/7!7! = between consecutive sessions, for a total of 14 sessions. Only
3,432]. Not all of these orders are desirable, however. For one condition was given during a session, and each condition
example, one order involves seven Condition A sessions occurred seven times. As described earlier, the conditions
followed by seven Condition B sessions. This would essen- were alternated in a semirandomized fashion with the stipu-
tially result in an AB design, and possible order effects could lation that no condition was presented during more than
make the results difficult to interpret. To ensure greater three consecutive sessions. Randomizing the conditions
alternation of conditions, we restricted the design so that controlled for order effects; providing each condition on a
randomization of conditions resulted in no more than three different day with at least 1 day between sessions should have
consecutive sessions of the same condition. This restriction reduced carryover effects (Barlow et al., 2009).
resulted in 1,972 possible random orders of seven sessions Fifteen elicited prompts were presented during the
per condition. context of a 20- to 30-min play session. Other intervention
studies used (a) a range of 10–20 prompts to imitate, (b) adult
models, or (c) expansions of child comments (cf. Loeb &
Experimental Procedure Armstrong, 2001; Scherer & Olswang, 1989; Willer, 1974).
Assigning a different semantic relation to each condi- A study by Fey, Yoder, Warren, and Bredin-Oja (2013) had a
tion would require that the two semantic relations be at the targeted rate of 60 prompts in 60 min; however, the prompts
same developmental level and be matched perfectly for degree used in their study ranged from a nonintrusive time delay to
of difficulty; however, it is difficult to ascertain that the dif- a more intrusive prompt to imitate. Because the only prompt
ficulty between any two semantic relations is equivalent for a used in this study was the intrusive elicited imitation prompt,
given child. Therefore, after identifying two target relations a maximum of 15 trials per session was imposed to prevent
that were produced but not mastered, one randomly selected child fatigue and to enhance compliance. Results from a pilot
relation served as a target in both the grammatical and tele- study indicated that 15 prompts in a 20- to 30-min session
graphic conditions until the child demonstrated mastery-level yielded sufficient responses from the child without causing
performance of the semantic relation. Mastery level was aversion to the prompts.
defined as 13 out of 15 multiword responses to prompts The sessions adhered to the basic principles of enhanced
to imitate in two consecutive sessions. If a child produced milieu teaching; specifically, each testing episode was made
13 utterances containing the target semantic relation in re- relevant by arranging the play context to increase the likeli-
sponse to the prompt during two consecutive sessions, a hood that the child would need to produce utterances contain-
second semantic relation was introduced as the target in both ing the target relation and by following the child’s attentional
conditions. This graduation to a second goal was necessary lead (Hancock & Kaiser, 2006). For example, the target se-
for only one child, identified as P5. mantic relation action + object may have consisted of specific
To minimize carryover effects, specific examples of prompts to imitate push (the) button, open (the) box, kick (the)
the target semantic relation were kept different for the two ball, and so on. Each specific prompt contained lexical items that
conditions. For example, the specific imitative prompt of Say, were known by the child as evidenced by the language sam-
duck walking occurred only in the telegraphic condition. ple, CDI:WS, and observations during previous experimental
condition. This resulted in a nonsignificant test statistic of not occur with future participants, we saw this child for an
1.57, which has a p value of .35. There was also no statistically additional six sessions to pilot the use of corrective prompts.
significant difference in the number of times P2 provided no Recall that a corrective prompt is an immediate and identical
response to a prompt between conditions (Mgrammatical = 4.0, second prompt to be used only if the child fails to produce
Mtelegraphic = 4.29; |A–B| = 0.29; p = .89). a target response to the first prompt—in this case, a two-word
Figure 2 also illustrates P2’s use of grammatical func- semantic relation. The two conditions were alternated, with
tors in his imitation of semantic relations. Visual inspection each condition occurring three times. The only change in pro-
shows clearly that P2 imitated at least a small number of these cedure was the use of corrective prompts, which were included
forms, but these productions occurred only in the grammat- in each of the total 15 prompts given per session.
ical condition. P2 produced a mean of 1.86 responses that Figure 4 displays the pattern of responses that con-
contained a function word (range = 0–5, SD = 1.95) in the tained at least a target semantic relation for these six sessions.
grammatical condition and a mean of 0 responses that con- The mean number of responses is identical for both conditions
tained a function word (range = 0, SD = 0) in the telegraphic (M = 8.67), and the ranges of scores are similar (i.e., gram-
condition, resulting in a significant test statistic of 1.86, which matical range = 5–12, telegraphic range = 6–12). There is no
has a p of .025. clear divergence of the lines, indicating that both conditions
were similarly effective. A randomization test could not be
Data for P3 completed on these data because the order of sessions was not
randomly assigned. Regardless, the procedure of corrective
P3 rarely responded to the prompts for her target of prompts was judged to be effective in eliciting multiword
agent (patient) + action by producing a semantic relation. responses and, therefore, was used with the final two partic-
Examples of imitative prompts for P3 are Say: The dog is ipants when necessary.
sleeping and Say: Bear driving. Instead, she frequently re-
sponded with just a single word regardless of the condition. Figure 4. Participant P3’s number of responses to corrective prompts
The numbers of responses containing a semantic relation are that contained a target semantic relation.
displayed in Figure 3. Despite the infrequent production of
her target semantic relation, the lines clearly overlap, indi-
cating no difference in her level of responding for either
condition. P3 had a mean of 0.29 responses that contained
a semantic relation (range = 0–1, SD = 0.49) in the gram-
matical condition and a mean of 0.57 responses (range = 0–1,
SD = 0.53) in the telegraphic condition. The resulting non-
significant test statistic is 0.29 ( p = .54). There was also no
statistically significant difference in the number of times
P3 provided no response to a prompt between conditions
(Mgrammatical = 5.43, Mtelegraphic = 5.86; |A–B| = 0.43; p = .75).
As Figure 3 also shows, P3 did not produce any function
words in response to the prompts in either condition.
As noted above, P3 frequently produced a single word
in response to the prompts to imitate. To ensure that this did
Data for P5 introduced. The pattern resembles the previous four cases.
There is no clear divergence of lines, demonstrating that P5’s
P5 began the experimental protocol with the semantic
level of responding did not differ across conditions. In addi-
relation of action + object as his target. During the fourth
tion, both lines trend downward for the final two sessions,
and fifth sessions, he imitated 13 out of 15 prompts, meeting
indicating that carryover does not appear to account for
the ceiling criterion for a target. Therefore, his second target
the similarity of responses. P5 had a mean of 6.0 responses
semantic relation, agent (patient) + action, was used for all
(range = 2–13, SD = 3.61) in the grammatical condition and a
remaining sessions. Examples of imitative prompts for this
mean of 6.14 responses (range = 2–13, SD = 4.78) in the
child are Say: Throw the ball [action + object] or Say: The fish
telegraphic condition. This resulted in a nonsignificant test
is swimming [agent (patient) + action] and Say: Blow fan [action
statistic of 0.14 with a p of 1.0. As with the previous four
+ object] or Say: Bug crawling [agent (patient) + action] for
participants, there was no statistically significant difference in
the grammatical and telegraphic conditions, respectively.
the number of times P5 provided no response to a prompt
After changing goals, P5’s level of correct responding im-
between conditions (Mgrammatical = 5.43, Mtelegraphic = 7.43;
mediately dropped far below the ceiling, indicating that
|A–B| = 2.0; p = .25).
agent (patient) + action was a more appropriate target.
P5 had a significant difference for the number of re-
Despite this change of targets, P5 responded similarly in both
sponses that contained a function word as well as the target
conditions. Figure 6 displays his pattern of responses; the
semantic relation. These results are also displayed in Figure 6.
vertical line at Session 6 indicates when the second target was
When action + object was the target semantic relation, P5
frequently imitated the function word, but only in the gram-
matical condition. After the ceiling criterion was met and
Figure 5. Participant P4’s number of responses containing a target the second target was introduced, P5 imitated fewer func-
semantic relation and number of responses also containing a function
word. tion words, but, again, these productions occurred only
in the grammatical condition. P5 produced a mean of 4.43
responses that contained a function word (range = 0–12,
SD = 3.95) in the grammatical condition and a mean of 0
responses that contained a function word (range = 0, SD = 0)
in the telegraphic condition, resulting in a significant test
statistic of 4.43, with a p of .008.
Discussion
This investigation sought to determine whether chil-
dren with language impairment who are just beginning to
combine words to express semantic relations produce utter-
ances containing target semantic relations more reliably in
response to prompts to imitate that are telegraphic than to
prompts that are grammatically complete. Our single-case