M ultiple regression of orthogonal polynom ials created from the w ord presentation rates (excluding rest condition) on adjusted BO LD contrast values from right and left auditory cortex voxels o f peak activation
1
First Left 537.851 R ight 612.424 Left <0.0001 R ight <0.0001 Second 38.707 117.260 <0.0001 <0.0001 Third 18.201 5.625 <0.0001 0.180 Fourth 5.421 0.175 0.202 0.6762Neural correlates o f selective attention
8.3 Discussion
8.3.1.
Stimulus dependency of rCBF measurements
These findings confirm that stim ulus presentation rate determ ines regional cerebral blood flow in prim ary auditory cortex. D ependency o f rCBF on stim ulus presentation rate has been show n previously for prim ary visual cortex (Fox and R aichle, 1984; F ox and R aichle, 1985) and prim ary auditory cortex (Frith and Friston, 1996; Price et a i , 1992). The nature o f the response is o f som e interest. M easurem ent o f rC B F by the Hj'^O rad io tracer technique requires accum ulation o f counts over a scanning w indow o f 90 seconds. The rC B F m easurem ents therefore represent an integration over tim e o f the h aem o d y n am ic resp o n se to neural activity w ithin the scan n in g w indow . If the haem odynam ic response accom panying each stimulus repetition is uniform , increases in stim u lu s presentation rate w ill resu lt in a linear in crease in the m ean in teg rated haem odynam ic activity. These data show that blood flow in auditory cortex show s an alm ost exact linear dependency on rate, in accord with previous w ork (Price et al., 1992). H ow ever, when sim ulated EPI sounds were presented alone, the rC B F was substantially low er than one w ould expect given the linear relationship betw een w ord presentation rate and rCBF. This nonlinearity may have its origins in the effects o f selective attention or arousal on auditory cortex. Technical lim itations w ith P E T scanning require that the auditory stim ulus com m ence som e seconds before the scanning w indow . T he subject therefore becam e aw are before the start of data acquisition w hether w ords w ere to be presented, and may have turned their attention elsew here for the duration o f the scan. In a sim ilar fashion, for those conditions w here words w ere presented the presence o f loud distracting background noise may have forced the subject to attend m ore intently than in the study o f Price and colleagues (1992) w here no background noise was present. In addition to the linear responses in auditory cortex. P rice et al show ed a nonlinear, stepw ise responses in left posterior superior tem poral gyrus (W ernicke’s area). H ow ever the d ata shown in F igure 8-2 com es from an area located w ell aw ay from p o sterio r superior tem poral gyrus, so this is not a plausible explanation for the present results. The difference between the rest condition and those conditions w here w ords w ere presented was an unexpected finding of this study and requires further investigation. A practical
Neural correlates o f selective attention
im plication o f this observation is that rC B F m easurem ents m ade w hen no stim uli are presented should be interpreted with caution.
8.3.2.
Stimulus dependency of BOLD measurements
T hese results show that there is a nonlinear, saturable relationship betw een stim ulus presentation rate and BO LD contrast m easurem ent. This finding is consistent w ith the nonlinear dependency o f B O LD contrast on rate o f presentation o f phonem es seen by B inder and colleagues (1994). They studied a phonem e discrim ination task w ith echo p la n ar im aging o f tw o slices through tem poral cortex (T R 3 secs) and sh ow ed a m onotonie nonlinear dependency o f signal on stim ulus presentation rate betw een rest and 2.5Hz. The present findings extend these results to passive w ord listening. The range of w ord presentation frequencies used in this study is com parable to that found in everyday conversation. T here is therefore no reason to suggest th at the saturable relation ship betw een w ord presentation rate and BO LD contrast is consequent on w ord presentation frequencies outside the normal physiological range.
8.3.3.
Why is there a difference between the BOLD and rCBF measurements?
T he difference betw een the PET and fM R I data, collected in the sam e subject under the sam e conditions, is striking. The rigorous and unique approach taken to reproducing the aco u stic ch aracteristics o f the tw o im aging env iro nm ents, w ith acq u isitio n o f an extrem ely large dataset in a single subject, allow s inferences to be m ade about the sp ecific n ature o f the relatio n sh ip betw een rC B F and B O L D m easu rem en ts. T he differences observed by previous investigators betw een PET and fM RI auditory studies are no t due to differences in paradigm , subject or stim ulus env iron m ent b ut reflect fu n d am en tally d ifferen t stim ulus depend en cies o f PET rC B F and B O L D co n tra st m easurem ents. The PET d ata illu strate that, for the co nd ition s w here w ords w ere presented, rCBF increases linearly with stim ulus rate. Therefore, the nonlinearity o f the fM R I data over the sam e range can be unam biguously attributed to factors o ther than cereb ral b lood flow . In this section factors other than cereb ral b lo od flow w ill be d iscu ssed , th a t m ay be rele v an t to the o b serv ed rela tio n sh ip b etw een stim u lu s presentation rate and BO LD contrast.
B O L D m easurem ents are contingent on changes in the concentration o f param agnetic deoxyhaem oglobin, an endogenous contrast agent. As a consequence there are tw o logical
N eural correlates o f selective attention
N B O LD m easurem ents are contingent on changes in the concentration o f param agnetic \ deoxyhaem oglobin, an endogenous contrast agent. As a consequence there are tw o logical possibilities for the observed relationship betw een stim ulus presentation rate and BO LD co n tra st. E ith er the rela tio n sh ip betw een B O L D c o n tra st and d eo x y h aem o g lo b in c o n c e n tra tio n , or th a t betw een w o rd p re se n ta tio n rate an d d e o x y h a e m o g lo b in