We have been doing a lot of research of Language Style Matching at the group level recently. One of the interesting aspects of that work is trying to determine the best way to capture mimicry (its easy when there’s only two people talking!). There are several methods. One of the most simple and elegant was proposed by Amy Gonzales in her excellent paper on small group dynamics.  You should read her paper for the details, but basically she considers every message as mimicking the group average.

One plus of Amy’s approach is that it can be easily calculated direct from LIWC output.  Here’s the SPSS syntax we’ve been using

AGGREGATE
/OUTFILE=* MODE=ADDVARIABLES OVERWRITEVARS=YES
/BREAK=Group
/SUMadverb=SUM(adverb)
/SUMarticle=SUM(article)
/SUMauxverb=SUM(auxverb)
/SUMconj=SUM(conj)
/SUMipron=SUM(ipron)
/SUMnegate=SUM(negate)
/SUMppron=SUM(ppron)
/SUMpreps=SUM(preps)
/SUMquant=SUM(quant)
/SUMpeople=NU(Group).

COMPUTE GROUPadverb=((SUMadverb – adverb) / (SUMpeople – 1)).
COMPUTE GROUParticle=((SUMarticle – article) / (SUMpeople – 1)).
COMPUTE GROUPauxverb=((SUMauxverb – auxverb) / (SUMpeople – 1)).
COMPUTE GROUPconj=((SUMconj – conj) / (SUMpeople – 1)).
COMPUTE GROUPipron=((SUMipron – ipron) / (SUMpeople – 1)).
COMPUTE GROUPnegate=((SUMnegate – negate) / (SUMpeople – 1)).
COMPUTE GROUPppron=((SUMppron – ppron) / (SUMpeople – 1)).
COMPUTE GROUPpreps=((SUMpreps – preps) / (SUMpeople – 1)).
COMPUTE GROUPquant=((SUMquant – quant) / (SUMpeople – 1)).

COMPUTE LSMadverb = (1 – (ABS(adverb – GROUPadverb) / (adverb + GROUPadverb))).
COMPUTE LSMarticle = (1 – (ABS(article – GROUParticle) / (article + GROUParticle))).
COMPUTE LSMauxverb = (1 – (ABS(auxverb – GROUPauxverb) / (auxverb + GROUPauxverb))).
COMPUTE LSMconj = (1 – (ABS(conj – GROUPconj) / (conj + GROUPconj))).
COMPUTE LSMipron = (1 – (ABS(ipron – GROUPipron) / (ipron + GROUPipron))).
COMPUTE LSMnegate = (1 – (ABS(negate – GROUPnegate) / (negate+ GROUPnegate))).
COMPUTE LSMppron = (1 – (ABS(ppron – GROUPppron) / (ppron + GROUPppron))).
COMPUTE LSMpreps = (1 – (ABS(preps – GROUPpreps) / (preps + GROUPpreps))).
COMPUTE LSMquant = (1 – (ABS(quant – GROUPquant) / (quant + GROUPquant))).

DATASET ACTIVATE DataSet1.
DATASET DECLARE GroupLSM.
AGGREGATE
/OUTFILE= ‘GroupLSM’
/BREAK=Group
/GLSMadverb=MEAN(LSMadverb)
/GLSMarticle=MEAN(LSMarticle)
/GLSMauxverb=MEAN(LSMauxverb)
/GLSMconj=MEAN(LSMconj)
/GLSMipron=MEAN(LSMipron)
/GLSMnegate=MEAN(LSMnegate)
/GLSMppron=MEAN(LSMppron)
/GLSMpreps=MEAN(LSMpreps)
/GLSMquant=MEAN(LSMquant).

DATASET ACTIVATE GroupLSM.
COMPUTE GLSMTotal= MEAN(GLSMadverb,GLSMarticle,GLSMauxverb,GLSMconj,GLSMipron,GLSMnegate,GLSMppron,GLSMpreps,GLSMquant).

 

 

Read more on Language Style Matching here:

  1. Richardson, B., Taylor, P. J., Snook, B., Conchie, S. M., & Bennell, C. (in press). Language style matching and confessions in police interrogations. Law and Human Behavior. doi:10.1037/lhb0000077
  2. Taylor, P. J., Dando, C., Ormerod, T., Ball, L., Jenkins, M., Sandham, A., & Menacere, T. (2013). Detecting insider threats to organizations through language change. Law and Human Behavior, 37, 267-275.
  3. Taylor, P. J., & Thomas, S. (2008). Linguistic style matching and negotiation outcome. Negotiation and Conflict Management Research, 1, 263-281.