The Lecture Circuit
 
 
We now move on to the concrete task of collecting data. To remind ourselves: the purpose of collecting the data will be to set up a system for demonstrating whether our treatment is proving effective for our client. As such, any measures will be individualized for the client at hand and will thus be nonstandardized and criterion-referenced. An important word in the accountability process is probe. The word probe can be a verb or a noun and is, of course, used in English beyond the topic we are discussing here. As a general verb, it means to explore, investigate, or poke around into something to see what you can learn (The captain said, “We will probe the planet’s surface for fuel.”) As a noun, it refers to a device that will allow you to explore (The space module used a giant probe to collect samples.) In data collection, to probe is to sample behaviors in a nonstandardized way (the verb) or the actual measure created to sample a specific behavior (the noun). We probe (explore) communication behaviors by using probes (the measures we construct). We must consider several factors in creating probes. These are: 1) What specific behavior do we want to measure? 2) At what linguistic level will the behavior be produced? 3) What degree of structure will be imposed? 4) In what communicative event will the target behavior occur? 5) How many opportunities of the target are required? Specific behavior. Probes allow us to zero in on behaviors known to be of interest. At the start of treatment, we focus on behaviors likely to be treatment targets, and then later, we probe those behaviors again to measure change. Because of our goal to demonstrate that our treatment is working, we will also probe behaviors beyond those being targets, so we can compare treated versus untreated behaviors. We must be sure we know what behavior we want so that we can construct a probe that will elicit it. Thus we must know the behavior (the particular phoneme, morpheme, or sentence type) and the specific forms we wish to sample. The forms sampled can be as narrow or broad as the clinician wishes (e.g., /s/ in word-initial position vs. in all word positions; /r/ in alveolar blends vs. singletons and blends; subject pronouns he and she vs. all pronouns; what and where questions vs. all Wh- questions). Linguistic level: Linguistic level (sometimes called linguistic unit, linguistic complexity, or response level) describes the language unit in which the behavior occurs. A standard hierarchy of linguistic levels is word, phrase, sentence, and connected speech. Some behaviors, such as phonemes, may occur in even smaller linguistic levels, such as in a syllable or in isolation. In addition, connected speech may in reality contain multiple linguistic units—some phrases, some sentences, some individual words. Typically, the lower linguistic levels of syllable, word, phrase, and sentence are sampled as individual units (one unit as a response to each stimulus, such as one word at a time or one sentence at a time), and the higher linguistic level of connected speech involves a spontaneous, running series of language units rather than a single unit at a time. One may also choose to create a linguistic level that is harder than the individual sentence but less complex than connected speech, such as requiring two-sentence responses. Probing multiple linguistic levels helps a clinician develop a clear picture of a client’s strengths and weaknesses. It is common, for example, to sample target phonemes in both words and connected speech, for a comparison in accuracy between the simpler single word and the more complex connected speech. By probing systematically along the hierarchy of linguistic levels, a clinician can identify where a client moves from skill proficiency to breakdown. Of course, for a client who has not yet begun learning a specific target behavior, low linguistic levels will reveal the absence of the behavior, and probing of higher linguistic levels will not be necessary. Degree of structure. Probes can be highly structured or loosely structured. I prefer “loosely structured” (or “low structured”) to “unstructured” to reflect the reality that we are planning the probe on some level to make sure we can sample the target behavior. Although we can, in fact, implement a truly unstructured probe by walking into a situation and simply observing the natural communication that occurs, I am focusing here on probes that we create with a purpose in mind, that is, probes to elicit specific behaviors. Probes with a high amount of structure have certain advantages. By imposing structure, the clinician increases the chance of eliciting the desired behavior. Structure also makes it possible for the probe to be administered quickly, thereby taking less therapy time. Here are some examples of high-structured probes: • To sample target phoneme /s/ in word-initial position, show the client 10 pictures that illustrate items beginning with /s/. Elicit the target by asking, “What is this?” • To sample target past tense, show the client a doll and make the doll enact different actions (e.g., have the doll jump, then stop). Elicit the target by saying, “What did this boy just do? He ___” (client fills in the blank). • To sample the target comprehension of prepositions, have a dog and several containers that allow placement of the dog in, under, on, behind, etc. Elicit the target by saying, “put the dog in/on/under the box” (note that the target in this case is an action that shows the client understood the preposition) • To sample the target produces sequences of ideas, present the child with 3-card picture sequences and have him describe the small story depicted. For an older client, show the client a newspaper article and ask him to explain it to you. As is obvious, the high-structured probes do not resemble natural communication. The imposition of high structure dilutes the communicative value, and the client may be aware of the “test-like” nature of the situation. Loosely structured probes are less contrived and thus bear more similarity to natural communication. Because they are less controlled by the clinician, they may not elicit the target as predictably and may take more time to implement. Examples of loosely structured probes (using the same target behaviors as named above) are: • To sample the target phoneme /s/, encourage the child to look at a book and talk about what s/he sees. Select a book that has pictures of words with /s/. • To sample the target past tense, give the child a variety of toys including people. Hope the child creates actions so that you can (as naturally as possible) say, “oh my! What just happened?” • To sample the target comprehension of prepositions, give the child toys that lend themselves to spatial arrangements. Intermittently give the child directions that fit in with play but are not completely predictable. You can also intersperse some instructions that involve prepositions throughout a session. • To sample the target produces sequences of ideas, have the client look at a book with you, and then have him or her tell you what he liked about the book. For an older client, use a conversational format to express interest in a topic the client is knowledgeable in—such as computers games—and ask the client’s opinion. Note that some behaviors are more readily sampled in loosely structured probes. For example, because pragmatic behaviors relate to communicative function, you need a sufficiently realistic communicative event—by definition likely to be more loose in structure. The opportunities for the behavior to occur depend on part on the natural flow of events rather than on contrived events created by the clinician. In this way, degree of structure interacts with nature of communicative event. Communicative event. Highly structured probes often occur in a drill-like structure, where the clinician presents a series of picture or printed stimuli and the client gives a series of responses. However, clinicians also use games to create structured probes with a more natural feel. Examples are eliciting single words with phoneme targets in Bingo; having the client throw a ball at pictures on the wall and then saying a word, phrase, or sentence to describe the target; creating sample sentences for a client to say in a game (Checkers, Sorry, Don’t Break the Ice), with the target embedded in the sentence (such as “my turn, your turn” or “I choose ___” or “don’t make it crack!”) Clinicians also create probes that simulate real events, such as (for an older client) having a client make a grocery list (to sample word retrieval) or having a client make phone calls to stores (to sample fluency). Finally, clinicians can use books for probes when the book allows a particular target to be elicited repeatedly. An example might be using a counting book to elicit the plural morpheme or a simple action story to elicit verbs. One can thus preserve some structure—thereby enabling specific sampling—while still making the communicative event more natural. In general, highly structured probes that are drill-like are sufficient in early stages of treatment when a client is new to a target, but as the client progresses through treatment, some more loosely structured probes will be necessary to demonstrate adequate generalization to natural communication. Number of opportunities. An opportunity is simply the chance to use the target response. The more structured the probe, the easier it is to make the opportunity obligatory (person has to use the target response). Because probes are nonstandardized and are individualized to the needs of the client, the number of opportunities for a target will vary. Keep in mind that the probe is designed to sample the behavior sufficiently for a realistic indicator of the client’s skill. Too few opportunities may fail to reveal variability according to phonetic or linguistic context; however, too many opportunities will make the probe unwieldy and impractical (especially for younger clients). For any given target, I consider 10 opportunities the maximum, and I would more likely use 5 opportunities when probing more than three different target behaviors. Keep in mind that probes are used at different stages in the treatment process. Before the onset of treatment, when the probes function as baseline, a client may well have few if any correct responses, and completing the probe may seem silly. However, as the client progresses in treatment, correct responses should start to emerge, and multiple opportunities will provide a better chance to observe the learning curve. Comparison to the original probe will be necessary for documenting the benefits of treatment, so completing the original probe even when there are no correct responses will be worth the effort. As shown by this long entry, there are many things to think about in constructing probes. My recommendation to those of you fairly new to using probes for accountability documentation is to start with simple probes. You will learn as you go along which probes give you the best information. In the next entry, we will discuss when to implement the probes at different stages of treatment. Slowly but surely, we are arriving at the actual documentation of treatment progress!
Monday, October 27, 2008
Measurement II:  Creating Probes to Collect Data