Functions of Repetitive Talk to Dogs during Play: Control, Conversation, or Planning?

Robert W. Mitchell and Elizabeth Edmonson
Eastern Kentucky University

This study describes people’s repetitive talk when playing with dogs and explores 3 hypotheses about that talk. Each of 23 people played with 2 dogs (one familiar, one unfamiliar). Videorecorded participants spoke about 208 highly repetitive words per interaction. Of all words used, 8 accounted for more than 50%. Phrases most frequently used and repeated were "come on" and "come here." In decreasing order of frequency, sentences ranged from imperatives to attention-getting devices, declaratives about the dogs, and questions. Additional declaratives and talk for the dog rarely occurred. Data support the conclusion that repetitive talk to dogs during play, with some conversational aspects included, mostly attempts to control the dog. Little evidence exists for "on-line" planning in talk to dogs.

Human animals confront the world through language. It is hardly surprising that we address entities that neither understand nor respond: machines, images, gods, as well as human infants and nonhuman animals (Snow & Ferguson, 1977; Snow, de Blauw & van Roosmalen, 1979; Hirsh-Pasek & Treiman, 1982; Caporael, 1986; Robins, Sanders & Cahill, 1991; Guthrie, 1993). Sometimes the speaker’s talk intends to influence indirectly someone other than the entity (Schottman, 1993). Usually, however, the speaker intends to address the entity directly.

Some broad similarities are apparent in these linguistic displays, at least during intense "interaction." Most show extensive repetitiveness, whether over a brief period as in speech to infants (motherese), talk to a failing car, and questions to a companion animal about the companion’s wants or over a more extended period as in religious rituals and prayers to divine entities. Except for talk to gods, most sentences to nonhuman and/or nonlinguistic entities are short, two to five words in length (Snow et al., 1979; Hirsh-Pasek & Treiman, 1982; Caporael, 1986; Robins et al., 1991).

In many ways, people's talk to animals resembles their talk to human infants (Katcher, 1985; Katcher, Beck & Levine, 1989; Hirsh-Pasek & Treiman, 1982; Mitchell, in review), the most well-studied of all talk to living, nonlinguistic entities (Snow, 1972; Newport, Gleitman, & Gleitman, 1977; Masataka, 1992; Chanu & Marcos, 1994). People's talk to young infants, children, and dogs provides repetitive instances of simpler, grammatically accurate forms (Snow; Newport et al.; Hirsh-Pasek & Treiman; McConnell & Baylis, 1985; Rogers, Hart, & Boltz, 1993; Mitchell, in review). Talk to infants and domesticated animals, or between lovers, sometimes shares a speech register with distinct pitch and intonation, "baby talk" (Ruke-Dravina, 1978).

That people talk to their companion animals about the animals' desires and the owners' emotional problems is well known (Katcher, 1981; Stallones, Marx, Garrity, & Johnson, 1988). Not well studied, however, are the words and sentences people use in talking to animals. One study of six elderly people who were walking their dogs described their talk to dogs as repetitive, consisting mostly of imperatives with only a few questions. (Rogers, et al., l993) Another study of four women, individually talking to their dogs both in a waiting room and at home, described similarities between such "doggerel" and motherese. The study noted repetitiveness, a high frequency of grammatically acceptable utterances, a low mean length of utterance, and similar frequencies of tag questions (Hirsh-Pasek & Treiman, 1982). A third study of 23 people who played in two interactions with a different dog in each (one familiar, one unfamiliar) examined more closely the differences between motherese and talk to dogs. The study noted that talk to dogs had shorter sentences, more repetitions, more imperatives, and fewer questions and declaratives than motherese (Mitchell, in review). These 23 people are the subjects of the current study.

The social responsiveness of both dogs and infants might be thought responsible for any similarities between caregiver talk to infants and to dogs, including repetitiveness and short sentence length (Hirsh-Pasek & Treiman, 1982). This explanation is incomplete, however, because mothers do not universally talk to babies using repetitive short sentences (Harkness, 1976). People who talk in one way to infants do not necessarily talk in the same way to domesticated animals (Brown, 1977). People sometimes use ordinary speech to address companion animals (K. Allen, personal communication, October 6, 1998; Goffman, 1981). Indeed, people need not use language at all to interact effectively with dogs (McConnell, 1990). Why, then, do people talk repetitively to dogs?

This article examines people's talk to dogs during play to describe what people say and to discover the functions of their talk by comparing it to similar forms of speech. The article explores the function of repetitive talk by examining three explanations derived from studies of repetitive talk to or by young humans. Talkers are trying (a) to control and get the attention of a dog, (b) to engage in a "conversation" or a "feeling of effective communication" with a dog and (c) to plan aloud their activities with a dog. In addition, this article discusses briefly the concern that, in ostensively talking to a dog while playing in front of a videocamera, people are simply engaged in performances or indirect speech acts toward the videorecorder.


Control and Attention-Getting

Much talk to infants is remarkably repetitive (Kobashigawa, 1969; Snow, 1972; Newport et al., 1977; Joci´c, 1978, 1990; Masataka, 1992; Chanu & Marcus, 1994). For example, consider a mother’s talk to her three-month-old infant:

In this highly repetitive talk, not only do a single question ("Are you a good boy?") and its deformation ("Are you?") account for most sentences, but only 11 different words are used. Three words ("Are," "you," and "a") account for most of them. Much of such repetitive talk from caregiver to infant occurs during highly repetitive activities such as feeding, play, and games, which may support the repetitive language used with infants (Stern, Beebe, Jaffe, & Bennett, 1977; Sylvester-Bradley & Trevarthen, 1978).

Dog-human play, like most play, is also highly repetitive (Simpson, 1976; Mitchell & Thompson, 1991), and talk to dogs during play is, at times, strikingly similar to motherese. The following talk to a dog is taken from the current study: "You're a good dog. You're a good dog. You are. You are good. You are silly, but good. Yes you are. You are a silly but good dog." Once again, a single statement and its deformations and elaborations account for most of the sentences. Three words account for most of the eight used.

In talk to infants, short, repetitive sentences arise "for the purpose of here-and-now communication with a limited and inattentive listener" (Newport et al., 1977, p. 126). As infants mature, the control and attention-getting aspects of such communication become inherent aspects of motherese. Mothers commonly try to direct and control a child's actions and attention. During the child's second year, however, repeated imperatives, sometimes disguised as questions ("Why don't you do x?"), predominate (Newport, 1977; Newport et al., 1977, 1990). At this time, motherese is "best described in terms of conversational meanings (the topic of directing action) and adjustments to processing limitations (keep it short and sweet)" (Newport et al., 1977, p. 129). Mothers may repeat because of "the child's failure to comprehend" their directives (Snow, 1972, p. 563) or because of the child's failure to comply with them.

Speech concerned with control and attention getting is likely to predominate in talk to playing dogs. Gaining control is, in part, the purpose of dog-human play. (Mitchell & Thompson, 1991). The talk’s failure may account entirely for its repetitiveness, based on continued attempts to control. To get a dog’s attention as well as to gain greater control, people often speak in imperatives (median percentage of 57.5% of utterances) and (median percentage of 25.5% of utterances) repeat the dog's name (Rogers et al., 1993). Repeatedly yelling the same phrases at the players in a televised football game displays similar repetitiveness in the face of failure to control. Although this talk cannot affect the players, it seems to provide the viewer, who is usually telling the players what to do, with a sense of participatory control. Similarly, when teachers talk to bigger groups of young children, as compared to smaller groups or individual children, they produce "significantly more unvarying, complete self-repetitions and fewer with expansions." In addition they show a lower mean length of utterance (MLU): 3.2 for seven children versus 4.0 for three children and 3.8 for one child (Pellegrino & Scopesi, 1990, pp. 107, 109). These differences probably result because "the prime aim of a teacher when interacting with seven children under three [years of age] is to control their behavior and particularly to call for their attention" (Pellegrino & Scopesi, p. 112).


Conversation

People may limit the length of sentences and repeat the same words and sentences when interacting with infants as well as dogs "to create the feeling of effective communication" (Snow, de Blauw, & van Roosmalen, 1979, p. 287). Simply gaining an infant's attention can create this “feeling,” which need not be tied to effective communication. "The mothers who spend so much time talking to their babies contend that the babies cannot understand them" (Snow et al., 1979, p. 269).

Some dogs apparently can understand a great deal of speech (Warden & Warner, 1928), and people generally believe that their dog understands some words, such as names of toys and familiar people (Walton & McConocha, 1996). As with young infants, however, people usually do not believe that a companion animal completely understands their speech.

Although caregivers talk to infants and companion animals without expecting a verbal response, they still expect some sort of response (Sylvester-Bradley & Trevarthen, 1978). Mothers often use a conversational model with their infants that permits the infant to respond to questions with behavior. When the baby does not respond verbally, (Snow, 1977), the mother can then fill in and speak for the child. People sometimes approach their pets similarly (Katcher et al., 1989; Rogers et al., 1993; Arluke & Sanders, 1996). However, people may not fill in answers to their own questions for animals in the same way that they do for infants (Hirsh-Pasek & Treiman, 1982), perhaps because animals more frequently answer the verbal questions with actions. Repetitiveness and short sentence length may represent the caregiver’s imposition of a conversational model:


The mothers' attempts to maintain a conversation despite the inadequacies of their conversational partners account for the most striking characteristics of the maternal speech style -- its repetitiveness, the high frequency of questions (especially tag questions and post-completers like "Hmm?" which are . . . devices for passing a turn on), and the frequency of sequences . . . in which the mother takes both parts . . . (Snow, 1977, p. 13)

Repetition, taking the infant's part, and tag questions are attempts at conversational "repair," making the mother's stream of talk seem like a two-party conversation (Snow, 1977).


Planning in Speech to Oneself

Repetition is also present in some problem-solving speech to the self. Repetitiveness here derives from the problem-solving nature of an encounter rather than from attempts to control another or create a conversational feeling. Consider the following private speech from a child in her repeated efforts to obtain an object on a shelf above her: "It's very high. It's very high. I'll have to call Lybna so she can get it. You can't get it. It's very high. I can't get it. It's very high. My arm's tired. You can't get it" (Levina, 1981, p. 286).

In this self-discussion, a child describes her difficulty and her inability repeatedly to herself, explaining to herself her dilemma. Although sentences are short and repetitive, as with speech to infants and pets, they are declaratives, rather than imperatives or questions, and focus on the child's own actions rather than on those of another. The sentences appear to focus attention on the problem and may not have the conversational tone present in speech directed outward. Such speech to oneself occurs, though infrequently, during talk to infants, but it does not appear to be repetitive and has a longer sentence length (Snow, 1977, p. 4).


Hypotheses

If people approach their dogs with a conversational model, one would expect the talk to contain many questions such as tag and post-completers, declaratives about the dog, few imperatives, many attention-getting devices, and some speaking for the dog. If people are largely trying to control a dog, one would expect their talk to contain mostly imperatives, including imperatives disguised as questions, attention-getting devices with few questions and declaratives, and little speaking for the dog. Attention-getting devices support both hypotheses. Suppose, however, that people are simply talking to themselves while they play, treating the interaction as a problem-solving situation with only the appearance of addressing statements to the dog. Then one would expect many "I" statements and descriptions of the situation (including the dog and its actions), few imperatives and questions, and little speaking for the dog.


Methods

The data analyzed here derive from a study that examined the organization of dog-human play and the effects of familiarity upon this organization (Mitchell & Thompson, 1990, 1991). In the present analysis, people's speech when playing with a familiar and with an unfamiliar dog was transcribed and the frequencies of types of words and sentences computed.
Participants

Participants

A total of 24 dogs and their owners, solicited through advertisement and word of mouth, were videotaped during dyadic interspecific play, resulting in 48 interactions, of which 46 were transcribed in the present study. Each person and dog played with a familiar and an unfamiliar partner. If person A played with his or her own dog and with person B's dog, B played with A's dog as well as B's own dog. The four resulting interactions comprised an "AB group." In addition, order of interaction for familiar and unfamiliar pairings was varied. In this way, the activities of people playing with a familiar and an unfamiliar dog could be examined.
People’s mean age was 30 years, ranging from 15 to 60 years; the mean age of the dogs was 6 years, ranging from 8 months to 14 years. The mean length of ownership of the dogs was approximately 4.3 years, ranging from 4 months to 13.9 years. Only four people had owned their dogs for less than 1.6 years, and these four sets made up two AB groups.

Procedure

Play interactions occurred in outside areas that were identical for all four interactions in each AB group. The videorecorder was visible during the play interactions; the owner of the dog was present at play between unfamiliars to avoid disrupting the dog by his or her absence. The owner of the dog was told not to interact during his or her dog's play with an unfamiliar human. When one person did interact, however, the data from this interaction between person A and dog B and from person A's interaction with her own dog were not used, leaving 46 pairs, 11 AB groups and 1 extra pair of interactions. Two rags and a piece of rope were saliently available for play, and each person was given a tennis ball and asked to begin the play interaction by throwing the ball. The person was told to play in whichever way, with whichever objects, for as long as he or she wished and to inform the videorecorder when finished. No further constraints were placed on the players. Not surprisingly, familiar dyads were more likely to play than were unfamiliar dyads (Mitchell & Thompson, 1990).

Coding, Reliability, and Data Analysis

Edmonson transcribed people's words directed to a dog, as well as words spoken for a dog, from the videotapes. Mitchell then checked the transcriptions and retranscribed them.Nonword vocal sounds directed to the dogs were added. Words clearly directed to the videorecorder or owner were also transcribed but are ignored in this analysis. The agreement in words between the two transcriptions was 80%. "Agreement" was in terms of near identity in sound and in meaning. "Give me" and "gimme" were accepted as agreeing about two words. After a new observer was trained to do reliability on one videorecorded interaction, she watched six videorecorded interactions selected at random. She was not given access to the transcriptions. The agreement in Mitchell’s was again 80%. Infrequently, what people said was not detectable. In these few instances, an attempt was made to detect how many words were said and when sentences began and ended. Approximately 152 unknown words from 65 sentences remain undeciphered.

We analyzed the transcriptions in several ways. We determined the total number of each type of word and categorized them as nouns, verbs, adjectives, adverbs, interjections, prepositions, conjunctions, or articles. We categorized verb phrases and questions into types. Following Newport (1977), we classified statements as (a) GAUs; utterances that sound like colloquial, conversational speech; (b) sentence fragments, phrases or constituents of GAUs in isolation; or (c) stock expressions, ungrammatical utterances, and grammatically unanalyzable utterances that included both nonword sounds and partially and wholly indecipherable utterances. Following Hirsh-Pasek & Treiman (1982), we classified the verb tenses (past, present, or other) of the GAUs (Mitchell, in review).

We examined the entire corpus for MLU and repetitions between subsequent utterances. We initially evaluated both MLU and repetitions using two methods (see Table 1) with utterances exactly as transcribed in Method 1. These utterances included nonword sounds like whistles, "Aw," "Whoa,” and excluded laughs or claps and undecipherable utterances (using sentences lengths as approximated). Method 2 completely removed these undecipherable sounds and utterances, including the word, “Oh,” from the corpus and sequencing.

Method 2 is similar to that espoused by Brown (1977, p. 54) for discerning MLU. Method 2 was apparently the method used by Rogers et al. (1993) to evaluate repetition between sentences. Unlike Brown, however, the present study did not count morphemes -- studies of motherese use words, rather than morphemes, to determine MLU of mother's speech to infants (Snow, 1972; Newport, 1977). Concatenates ("I'm," "can't," "where'd," "gimme”) and semimodals ("gonna," "wanna") were counted as two words. The phrase, "tug o' war," used eight times, was counted as one word. We also examined repetitiveness in the entire original corpus (sounds treated as words and unanalyzable utterances) by comparing each utterance to its preceding 10 (or fewer) utterances. Newport (1977) espoused this method, also apparently used by Hirsh-Pasek & Treiman (1982), who do not state explicitly how they evaluated for repetitions (Mitchell, in review.

For most sentences in which all words were detected, deciding where the sentence began and ended was based on intonation, pauses, and standard English usage. Most of 33 syntactically correct sentences that contained the phrase, "Come on," followed or preceded by a command, could also be viewed as two or three separate GAUs. Excluded from this group of sentences, however, was syntax exemplified by "Don't pay attention to them, pay attention to me.” Also excluded weresentences with tag questions, those that could be transformed into a GAU, and a sentence fragment or stock expression.

As either interpretation made little difference (a maximum decrease of 0.026 for MLU and of 0.014 for number of repetitions using either Method 1 or 2), we interpreted these as single sentences. By contrast, Snow (1972) scored such run-on sentences as two or more utterances, and Rogers et al. (1993) appeared (from a few examples given) to have had an abundance of run-on sentences.

To look more closely at the different types of utterances in relation to the hypotheses, we reexamined the entire original corpus: the function of talk to dogs, use of imperatives, questions, and declaratives (including talk for dogs). We looked at attention-getting devices (calling the dog's name, words such as "Hey," and sounds such as "T-t-t-t-t-t." We also examined imperatives disguised as questions that were not "a request for information, but a directive" (Newport, 1977, p. 193), the topic of declaratives (the self, the dog), tag questions ("You like the ball, don't you?"), post-completers ("Don't you?"), and other salient sentence-types.


Results

Words and Phrases

Table 1 presents the frequencies of the most commonly used words and of their repetition. Table 2 presents the grammatical categorizations of these words. Tables 3, 4, and 5 present the most common imperatives, declaratives, and questions. Table 6 presents the frequencies of imperatives, questions, attention-getting devices, and declaratives. Table 7 presents MLU, repetition rate, frequencies of types of GAUs, and comparisons to the same aspects in other studies of talk to dogs.

People generally spoke a lot to the dog they were playing with. We discerned a total of 9,046 words (based on 384 types of words). We detected 351 nonlinguistic expressions: whistles, "Ow," "Oh-ho-ho," and "At-at-at-at." When the nonlinguistic expressions and the (approximately) 152 undecipherable words were included as words, people used, on average, about 208 words per interaction (range from 0 to 544). There were 3,767 utterances, 2,504 of which were grammatically acceptable and 63, sentence fragments ("To me," "Paw," "Against the wall?"). A total of 859 were stock expressions ("Hey," "Kasha," "Good girl," "Thank you"), Two were ungrammatical ("You're bored easy, don't you?" and “Them was my fingers.”), and 339 were unanalyzable as any of the above (“Ooo-oo-oo-oo-oo-oo,” “Whoa,” “Watch this ---”).

The MLU ranged from 2.54 (using Method 1) to 2.62 (using Method 2) to 3.07 (using GAUs and sentence fragments only) to 3.09 (using GAUs only). A total of 3,716 utterances (including 288 unanalyzable utterances) could be coded as imperatives, questions, attention-getting devices, or declaratives (only 51 unanalyzable utterances could not be coded). Almost all utterances in the corpus were addressed to the dog, but only a very few (six) were spoken as if for the dog.


Table 1. Overall frequency of most frequent words, and the frequency with which these words occur in consecutively repeated utterances

--------------------------------------------------------------------------------
 

Note: This list includes "gimme," "wanna," and "gonna" as "give me," "want to," and "going to." Any contraction (e.g., "isn't," "I'll," "don't") is counted as an instance of its first word (i.e., "is," "I," "do") and second word (i.e., "not," "will," "not"). Nonword sounds include imitations of dog's sounds, whistles, and other interjections, but do not include laughing or sounds made by hands and fingers.

--------------------------------------------------------------------------------

Using either Method 1 or Method 2 to determine repetitiveness led to approximately the same result (see Table 1): Many words were used repetitively. On average, people repeated from 0.79 words from one utterance to the next (using Method 1) to 0.88 words (using Method 2). When nonword sounds were ignored, the 29 word types in Table 1 account for 74% of the total decipherable words uttered. The four most frequently used words (“Come,” the dog's name, “on,” and “here”) account for 37%, the eight most frequently used words for 53%. Most people used the same words over and over again.

Not unexpectedly, the most frequently used (13%) word is “Come.” The most frequent sentence or phrase is “come on,” followed by “come here.” “Come” is also the word most frequently repeated, being present in the preceding sentence for more than 50% of the sentences in which it appeared. Because "on" and "here" are repeated in the next sentence for about 40% of the sentences in which they appear suggests that the phrases, "come on" and "come here," are repeated highly in consecutive sentences. Indeed, "come on" is repeated for just under 20% of all consecutively repeated utterances (18.7% using Method 1 and 19.1% using Method 2). "Come here" is repeated for just over 10% of all consecutively repeated utterances (11.7% using Method 1 and 12.5% using Method 2).

The second most frequently used word is the dog's name. Used to gain a dog's attention, it is uttered alone in 53% of its uses. The extensive use of "ball" indicates its frequent use as a play object.

The most frequently used word forms (see Table 2) are verbs (34%), nouns (28%), prepositions (12%), and adverbs (9%). Interjections (7%), included an extensive variety of nonword sounds, about one-third of which are imitations of dog sounds:

1. Ugh! sympathy;
2. Aw! excitement;
3. Yahoo! intense interest;
4. Ooooh! abrupt surprise; and
5. Woah! demarcation of a ball’s trajectory;

Additional interjections included “Wheeee!” with rising and falling intonations; “Shoo!” beginning when the ball is thrown straight up in the air; "At-at-at, nah-nah-nah!” a command for the dog to stop doing something; an articulated "Ha ha!” simulated laughter; and "Oh-ho-ho-ho!” engagement in the dog's actions.




--------------------------------------------------------------------------------
Table 2. Frequency of words used in play with dogs, sorted according to grammatical category, with words which occurred 20 times or more specified

Note: As in Table 1, semi_modals such as "wanna," and "gonna," as well as contractions such as "gimme," "isn't" and "you're," are counted as two words in this calculation. A number in parentheses represent the number of instances of the word directly preceding it that is included in total. Differences in frequencies between Tables 2 and subsequent tables usually occur because Table 2 concerns the frequencies of words, whereas subsequent tables concern the frequencies of utterances.
--------------------------------------------------------------------------------

Commonly used phrase types (see Tables 3 and 4) included "Go get it," "Go get the ball," "Get the ball," "Get it," "Good girl/boy/dog," "You're a good girl," "What a good girl you are," "Give me that ball," "Look," "Bring me the ball," "Bring it to me," "Drop the ball," and "Drop it." Although the word "No" is spoken several times, almost 25% of its uses occurred in one imperative in which it is repeated 16 times. The most frequently used questions (see Table 5) asked a dog if he or she wants something or wants to do something and "Ready?" or "Huh?” used either as tag questions or as post-completers.

--------------------------------------------------------------------------------
Table 3. Number of sentences in which frequently used verb phrases occurred as an imperative in an utterance or as part of an utterance
COME 1,213*

Notes: "DO" represents a present tense action verb; "x" represents a word or words; words in brackets are optional; italicized words are abstractions describing the types of words used in that position in the utterance. Note that this table concerns frequencies of utterances (which may contain the same word more than once), whereas Table 2 concerns frequencies of words. *"Come on" and "come here" occurred together in one sentence.

*Imperatives, declaratives, or questions are not mutually exclusive categories (as utterances can contain both, and imperatives can be disguised as questions), and deictic utterances can be any of these. Tag questions can be attached to other questions (e.g., "Where's the ball, huh?"), and are included in the percentages for Question forms total. Declaratives include the 6 utterances which talk for dogs.
--------------------------------------------------------------------------------

Most utterances repeated some part of the previous utterances. Using Method 1, 42.8% of utterances contained at least part of the immediately preceding utterance. Using Method 2, 47.1% of utterances contained at least part of the immediately preceding utterance. Using Newport's (1977) method, 68.5% of utterances contained at least part of an utterance from among the 10 (or fewer) utterances preceding it.

Evidence for Conversation, Control, or Self-Talk

Of the 3,716 utterances that could be distinguished as to sentence type (see Table 6), most (57%) are (or contained) imperatives. Most imperatives (55%) are (or included the phrases) "Come on" or "Come here." Other frequent imperatives employed the words "Go get," "Look," "Give," and "Get." The few nonword imperatives mostly implied "No": "At-at-at," "Na-na-na," and "Ugh!"
--------------------------------------------------------------------------------
Table 4. Number of sentences in which frequently used phrases occurred as a declarative in an utterance or as part of an utterance
 

Note. See notes for Table 3. Also, main categories in this table are not mutually exclusive, although subcategories are. Labels here are not preceded by either "good" or "you."
--------------------------------------------------------------------------------

Some questions seem to fit easily the criterion of being disguised imperatives, particularly those beginning with modals such as "can," "may," "could," "would," and "will" followed by the word "you." Also fitting easily into this category are imperatives having "Why don't you" before the imperative ("Why don't you get that ball again?") and imperatives with a questioning tone ("Give me the ball?” "Drop the ball?"). Other questions, however, are less clearly defined as disguised imperatives. Questions asking the dog about his or her wants ("Wanna play?", "You want the rope?") could be interpreted as disguised imperatives, but we viewed them as asking the dog about his or her wishes in accordance with how similar statements are interpreted in studies of motherese (Snow, 1977).

A few questions that asked if the dog might engage in particular actions ("Do you go after sticks, Chris?", "You play rope?"), why the dog didn't do something ("Hey, you don't bring it back?", "You don't want to play?"), or suggested possibilities for future action ("How about a stick?", "How about a tug o' war?", "Ball?") might also have been disguised imperatives. Similar questions addressed to a two-year-old child might indicate an implied command (Newport, 1977, p. 193). We viewed these questions as concerned with what the dog might do or expressing playful exasperation with the dog's inadequate actions. As such, these questions about a dog's actions or the self in relation to the dog can be viewed as attempts at problem solving.

Second most frequent is the category (30%) of attention-getting devices that frequently (72%) use the dog's name. The most frequent nonword attention-getting device is whistling, but another relatively frequent one is making dog-sounds, including growling ("Rrrr") or more conventionalized growling ("Woof," "Rrowrr," "Ruff," and "Rupp"), each alone or repeated several times. Both imperatives and attention-getting devices are attempts to control dogs, which seem to have been uppermost in people's concerns, but attention-getting devices can also serve a conversational function by drawing the dog's attention to the person as a part of the ongoing behavioral dialogue. Behavioral attention-getting devices such as showing objects (Mitchell & Thompson, 1991) often supplement linguistic attention-getting devices, calling the dog's name or saying "Hey"!

The next most frequent category of utterances, declaratives about the dog, comprised 15% of utterances. Most of these (65%) are reassurances toward the dog. Reassurances included the word "good" to characterize a dog or its actions. Other positive characterizations included "pretty," "sweetheart," "silly," "honeybun," as well as descriptions of the current actions: "Here you go," "There," "Oh yeah," "Yes," "Okay"; and nonword soothing sounds such as "Oooh," "Oh-ho-ho," and "Aw." A few declaratives suggested the beginning of problem solving in that they described or explained the problem of the dog's failure to play by describing the dog's mental state, behavior, or the dog's view of the objects available. "That one's boring," "This is baby stuff," "Oh, you're too pooped," "Don't like the ball," "You don't like to play fetch," "You're scared," "You don't wanna play," and "You're sweating". The dog is usually referred to as "you," although four times people used "we" to refer to the dog, such as "We're going to take a little rest", Only infrequently did people negatively characterize the dog ("dumb dog," "stupid"). Although not included in the corpus, people sometimes talked about the dog in the third person, usually describing the dog's actions ("He's looking for a way out") or failures to act ("He's not going to play"). In making these statements, people were trying to solve the problem of how to play with the dog, but they usually did not address these verbal descriptions and explanations of the problem to the dog.

Talk for the dog was a very infrequent declarative about the dog, with only six utterances. One older man specifically stated that he was speaking the (unfamiliar) dog's words by stating "He says" before the dog's (pretend) utterance. In this instance, the dog's thoughts are being described, indicating the dog's age in "dog years" as a reason for the dog's lack of interest in playing: "All right. Rest my weary bones. Ninety-eight and they want me to play ball." (Another sentence by this man, who might have been speaking for the dog, was not included in the corpus as he seemed primarily to be presenting his own view: "If you were ninety-eight, you wouldn't play ball either.") The other instance of speaking for a dog that indicated a dog's thinking was "She can't even reach me," which related to the speaker's inadequate toss of a ball. This woman spoke for dogs four other times (for two different dogs), all of which seemed directed to the person herself and implicated the dog's failure to play: "I belong to my mother," "Oh leave me alone," "Yeah, you come and get it," and "You can't just leave me alone."

Questions formed the next most frequent category of utterance (11%), and most of these (34%) concerned the dogs' psychological states, most usually those directly related to the dogs' actions (what dogs want to do). Some questions (30%) were tag questions ("You like the ball, don't you?” "You wish Ebony was here, huh?") or post-completers ("Don't you?", "Huh?", "Okay?"), which suggest a conversational element to the talk. Other questions (12%) were imperatives disguised as questions ("Can I have the ball?” "Gonna catch?” "Give me the ball?"). Less frequently, questions described a dog's actions or their implications ("No play?” "One more try?” "Are you calling it a night?” "You got a bally?") or asked a dog about the situation (usually in the form "Where's x?", "Where'd x go?", "What's x?", or "Who's x?"). People infrequently answered their questions to the dog with verbal responses. One woman, for example, hid a ball and asked "Where'd it go, Tope? Where'd it go? Where'd your ball go?” and then showed the ball saying, "There it is." More usually, people (and dogs) "answered" questions behaviorally. Only one use of "yeah," but no uses of "yes" or "no" were answers to questions, although some questions ("No?” "No play?") appeared to be answers with questioning intonation. That people question dogs indicates a conversational aspect to their talk.

Very infrequently people's utterances (3%) were declaratives about themselves, usually describing what they were doing or about to do -- an element of on-line planning. Most of these sentences started with "I," although two sentences started with "Mommy" to indicate self. Only twice did people describe their psychological states alone ("I'm tired"), although psychological characterization of themselves was present for some descriptions of the dog's mental states ("I can see where you're going to be a little shy with me, Tinker." "I know you're tired"). The infrequency of declarations about the self indicates that "on-line" planning is not a significant focus of talk to dogs.

People (1%) very infrequently simply named the objects ("Ball"), described the owner when playing with an unfamiliar dog ("Mommy's not here"), or described an object's action by itself, without reference to themselves or the dog ("Gonna be a big one", referring to the height of a thrown ball).

Indirect Speech Acts and Talk to Dogs as Performance

In some cases, sentences ostensively uttered to dogs were indirect speech acts. Telling a dog that he or she is tired or bored or emphasizing a dog's lack of activity and lack of desire to play indicated by indirect speech to the videorecorder that play was winding down or not starting. Still, people normally seem to talk to dogs when unaware that they are being observed. People self-report this tendency and commonly informed the videorecorder directly that the play was over.

People sometimes commented to themselves or to other people about a dog's actions ("She drops it right at your feet") or the dog's unplayfulness ("He's not going to play"). One woman whose golden retriever consistently failed to run after a ball said, as an aside, "Yes, this is Topaz and her human retriever." One man asked an unplayful dog's owner, "This dog's name's Chrissy?" After the owner responded "Yes," the man commented "Okay. It better be. Hope it's not Al or anything." Because such comments were not addressed to the dog, they are not analyzed in the corpus.

Although any interaction with a dog was a performance of sorts, in that people tried to reproduce their script of play with a dog, usually neither the behavioral interaction nor the talk seemed to be a performance. One man who had difficulty getting a dog to play, however, did perform "playing with a dog." He satirized an unfamiliar dog's inactivity by suggesting absurd "play" activities to the largely unplayful dog:


Now, let's play tug o' war. Come here Libby. Here, let's play tug o' war, Libby. Here. Here. Here you go. [The dog does not respond, so the man holds the two ends of the rope taut, one near the dog's mouth and the other away from the dog, and fakes strain:] Yee-ee-ee-ee-ee-ah-e-yah! Oh Libby, you won. Good dog. Good dog, Libby. Good dog. Oh, good Libby. Good Libby. Good. Good. Okay. Okay, Libby. Now, let's play hang Mark. Okay, here we go. [After wrapping rope around his neck, man fakes choking:] Ahhhwgg! [Dog moves to sitting position as man "chokes."] Oh, good dog, Libby. Good dog. Good dog. Good dog. Okay. Now, let's play sit. Sit Libby. [Dog lifts paw up, which man grabs and shakes.] Shake hands! Good Libby good. Good Libby. Good Libby. Okay. Now, [dog stands and walks away] let's play stand up. Let's play, let's play, let's play stand up. Let's play stand up, Libby. Come on. Let's play stand up. Okay, Libby. All right, Libby. Let's play "Gone with the wind." Come on, Libby. Let's play "Gone with the wind." Come on Libby. [Man sings theme song from "Gone with the wind":] Ah-Ah-ah-ah. She's back in the red earth of Tara. Look at her. Look at that emotion.

In this extensive parody of the play situation, the dog, although ostensively addressed, increasingly became an object in a pretend scenario. By the end of it, she was simply a prop, and the sentences were about her as a prop and not addressed to her (thus, the last four utterances were not included in the corpus). This sort of extended performance was unique.

Comparisons to Other Studies of Talk to Dogs

As can be observed in Table 7, the present study's findings are similar to those of two other studies on talk to dogs. In comparison, we observed more repetitiveness, a greater percentage of imperatives (and consequently lesser percentages of other sentence types), and about the same proportion of sentences containing the dog's name. Our MLU is similar to that of one study and less than that of another. This latter difference may derive from the other study’s more frequent use of run-on sentences in determining MLU. Any discrepancies may depend upon the different sample sizes, number of utterances, or contexts in which people talked to dogs.




--------------------------------------------------------------------------------
Table 5 Frequency of utterances in which these phrases occur frequently as questions or as part of an utterance