quantifying data

Scientifics,

the first major step in analysing your data is to quantify as many things as possible that will test your hypothesis e.g. if you are hypothesising that you will find defecit features, you will need to count how many hedges, empty adjectives etc. you find. These quantified findings will tell you what is significant (common/patterns of results) and worth exploring or unexpected and therefore worth exploring.

Set out a protocol for anything where it might be ambiguous what you are counting and not counting e.g. for interruptions you might want to establish that you are not going to count any back-channel agreement i.e. only count something as an interruption when it's a contradiction, agenda shift or the first person stops talking immediately or within a second of when the second participant speaks.

You might want to sub-divide e.g. count all interrogatives but keep them also as separate totals for open, closed, tag, rhetorical, prompt questions etc.

Keep them sub-divided also if you have multiple pieces of data so you can see if there are anomalies (surprising results, outliers etc.) in particular contexts e.g. if the participant's number of interrogatives decreases in one transcript, it would be a signal to look closely at why in that context fewer interrogatives were used.

Remember to make different pieces of data comparable by creating averages or percentages - it's no use pointing out that fewer interrogatives were used if it is a shorter transcript, so you should work it out as 'interrogatives as a percentage of turns taken' and acknowledge that, in small data samples, quite big percentage differences can result from just one or two extra/fewer examples... so discuss reliability and give the actual figures too e.g. in trascript one, participant A asks 40% (as a percentage of turns) interrogatives, however in transcript 2, she only asks 20%. This may be because her conversational partner in transcript 1 is female (and she uses a similar percentage of questions: 35%) and her partner in transcript 2 is male (and he uses a smiliar percentage of questions: 10%) which could be explained by Giles's theory of accommodation and by Tannen's orders vs proposals pairing from Difference theory, or it could be that that percentage of questions is idiolectal, as the difference in actual number of interrogatives over a similar length of transcript is just 1 in real terms. It may also be...

Once you have quantified anything you need to in order to decide how far your hypothesis is supported or contradicted, you can start doing close PEE anlaysis in context to explore why this might be the case according to any relevant theory and the GRAPE - be tentative and don't come to any firm conclusions. Ever. Never say proves or disproves. Aways acknowledge the limitations of the data.