• No results found

Development of a communicative spontaneous speech task

2.1.2 Referential communication tasks

Referential communication tasks can be used as an alternative to spontaneous conversation in the elicitation of speech. In a referential communication task, interlocutors are typically separated by a barrier, and they are given the task of identifying, describing or locating certain referents to each other. Thus, to enable the successful completion of the task, information must be exchanged between interlocutors, and both speaking and listening skills must be used in context by each participant, as in real communicative situations (Leinonen and Letts, 1997;

Lloyd, 2003; Markman and Makin, 1998; Yule, 1997). Referential communication tasks enable the researcher to retain some control over the speech produced, as the context and referents produced in the talk will be constrained by the task. These tasks also typically enable a higher event density than spontaneous conversation, as the referents in the task can be controlled by the experimenter to include ele-ments of interest. Critically for the current study, the researcher is aware of what each participant is trying to communicate, which is especially important when dealing with speakers who are less intelligible or who may be unable to produce certain phonemes accurately. The researcher also knows what the aim of the in-teraction is – trying to transmit information accurately to the interlocutor – thus enabling a measure of communicative success to be used to characterise the inter-actions. Although referential communication tasks are unlikely to elicit as natural

speech as spontaneous conversation, the speech produced is still spontaneous and communicative. Children are likely to encounter referential communication tasks in their everyday lives, for example in the classroom or in games played outside of school (Lloyd, 2003).

As seen in chapter 1, referential communication tasks have been used pre-viously to assess HI children’s communication skills, but the measures exam-ined have been limited to either the type of repair strategies or clarification requests used (e.g., Ibertsson, Hansson, M¨aki Torkko, Willstedt Svensson, and Sahl´en, 2009; Jeanes et al., 2000; Most, 2002; Sandgren et al., 2010), or the pro-duction and detection of ambiguity by participants (Arnold et al., 1999; Lloyd et al., 2005), with no known studies examining the acoustic-phonetic properties of speech produced by HI children in these tasks. In other research areas, referential communication tasks are often used to assess a speaker’s ability to prosodically disambiguate utterances (Ito and Speer, 2006; Niebuhr and Michaud, 2015).

Many referential communication tasks used in studies with children involve picture-array tasks, in which one of the participants (the ‘follower’) is given an array of simple pictures differing in critical attributes, and the interlocutor (the

‘leader’) is given one of the pictures to describe to the other. The follower then has to select the correct picture based on the leader’s description. This is the type of task used in, for example, the studies by Lloyd et al. (2005) and Lloyd et al. (1998) (see chapter 1). The Map Task (Anderson, Bader, Bard, Boyle, Doherty, Garrod, Isard, Kowtko, McAllister, Miller, Sotillo, Thompson, and Weinert, 1991; Brown, Anderson, Yule, and Shillcock, 1983) was created as a more complex alternative to picture-array tasks. The task leader is given a map with a route, and the task follower must draw the same route on their map based on the task leader’s description. Both the task leader and follower’s maps have ‘landmarks’ on them, which contain certain features of interest to the researcher – however, only some of these landmarks are shared by both interlocutors. The interlocutors need to navigate their route through the landmarks to achieve success in the task.

However, in both the picture-array tasks and the Map Task, interlocutors are given particular pre-defined roles of leader and follower in the conversation, and therefore, unlike natural speech, the elicited speech is likely to be less balanced between interlocutors and, in the Map Task, may mostly include instructions and

commands (Baker and Hazan, 2011).

One of the referential communication tasks used in the current study is the Di-apix task, which was originally created by Van Engen, Baese-Berk, Baker, Choi, Kim, and Bradlow (2010), and which was further developed by Baker and Hazan (2011) to enable the elicitation of spontaneous speech in a context in which both interlocutors can contribute equally and collaboratively to the conversation. In the task, a pair of interlocutors is given a different version of a picture-scene, and their aim is to find 12 differences between their pictures without seeing each others’ pictures. Because the picture-scenes are detailed, and contain small dif-ferences between pictures, the Diapix task elicits natural and fairly complex lan-guage from participants. The DiapixUK pictures contain a full set of 12 different picture-pairs which are of equal difficulty, and which therefore allow participants to complete several pictures both between and across conditions. The Diapix task has been used successfully in many recent studies, especially those investigating the global acoustic-phonetic properties of speech in adults (Hazan and Baker, 2011), children (Hazan et al., submitted; Pettinato et al., 2016), and second-language speakers (Granlund et al., 2012; Wester, Garc´ıa Lecumberri, and Cooke, 2014). The Diapix task is used in the current study due to its elicitation of fairly sophisticated syntax and vocabulary, and due to its relative difficulty as a task – the interlocutors are not given specific strategies or roles within the interaction, but must negotiate these between themselves. However, it could not be used to elicit segmental contrasts in speech – although the task was designed to elicit minimal pair keywords, which were included as objects in the pictures, Baker and Hazan (2011) demonstrated that the task did not reliably elicit sufficient numbers of these keywords for segmental contrast analysis.

Therefore, because one of the main aims of the current study is the elicitation of several different types of measures from NH and HI children’s peer interac-tions, including several repetitions of segmental contrasts to enable between- and within-category analyses to be done, an additional task was needed which would elicit frequent repetitions of minimal pair keywords in interaction. A few pre-vious referential communication tasks have been developed for similar purposes.

Sankowska et al. (2011) used a variant of the Map Task to elicit intrinsically long and short vowels in ADS, FDS and in noise. The target sounds were included

as street names that participants needed to use to find the correct route on the map. In Garnier et al.’s (2010) river task, both interlocutors were given a sheet of paper with 17 river names, which included target syllables. The task leader had to tell the interlocutor to connect each river to two others to make a route – the task follower then drew the correct route according to the leader’s directions.

However, although the task elicited communicative speech from the task leader, it did not require much interaction. The elicited speech was also highly constrained, therefore preventing a linguistic analysis from being done on the data.

The referential communication task developed in the current study was in-spired by the ‘SAME/TRAP’ task created by Hazan and Kim (2013). The aim of their task was to elicit consonant stop voicing (/b/-/p/) and place of artic-ulation (/b/-/v/-/d/) contrasts, as well as four point vowels, in an interactive task. In their task, each participant was given a 4-by-4 grid, with a coloured object and a letter in each square. Some of the letters corresponded to the target consonants, and the object in the square was an item beginning with the same letter as that in the square. The colours of the letters, on the other hand, were designed to elicit each of the point vowels. Participants’ squares were not identi-cal, and the pair of participants were to identify the squares which were identical (‘SAME’), and those in which the letter was the same but the object differed (‘TRAP’), without directly naming the object in the square. Therefore partici-pants produced sentences such as “Is your green V something you drive?” (Hazan and Kim, 2013, p.2). Although the task elicited fairly natural speech, and was successful in eliciting multiple repetitions of the target consonants and vowels, the nature of the target consonant elicitation prevents any minimal pair contrasts being used which occur elsewhere than in the syllable onset. The task may also be too complicated for HI children with language delays, as fairly sophisticated language skills are required to describe objects without naming them.