diff --git a/functions/01_TotalActivation/Temporal_TA/TA_Temporal_conv.m b/functions/01_TotalActivation/Temporal_TA/TA_Temporal_conv.m
index 58b779a..2d78ab0 100644
--- a/functions/01_TotalActivation/Temporal_TA/TA_Temporal_conv.m
+++ b/functions/01_TotalActivation/Temporal_TA/TA_Temporal_conv.m
@@ -1,74 +1,75 @@
 %% This function performs the temporal regularization part of total 
 % activation, voxel after voxel
 %
 % Inputs:
 % - TCN is the n_time_points x n_ret_voxels 2D matrix of data input to the
 % regularization
 % - param is a structure containing all TA-relevant parameters; here, we
 % will need the fields 'Dimension' (X, Y, Z and T sizes), 'NbrVoxels'
 % (number of voxels to consider for regularization), 'LambdaTempCoef' (used
 % to compute regularization coefficients)
 %
 % Outputs:
 % - TC_OUT is the n_time_points x n_ret_voxels 2D matrix of outputs from
 % the regularization step
 % - param is the updated structure with relevant parameters for TA; added
 % fields are 'LambdaTempFin' (vector with final regularization estimates
 % for each voxel), 'NoiseEstimateFin' (final noise estimate for each voxel)
 %
 % Implemented by Isik Karahanoglu, 28.11.2011
 function [TC_OUT,param] = TA_Temporal(TCN,param)
     
     % The output from the algorithm (time x voxels) is initialized as
     % a matrix of zeros
     TC_OUT = zeros(param.Dimension(4),param.NbrVoxels);
     
     % LambdaTemp contains the values of regularisation parameters for each
     % voxel; also set to zero for now
     param.LambdaTemp = zeros(param.NbrVoxels,1);
     
     % The noise estimation procedure uses a single scale
     % wavelet decomposition. Here Daubechies wavelets with 4 vanishing 
     % moments are used. The corresponding high pass filter is given by:
 
     g=[0    -0.12941    -0.22414     0.83652    -0.48296];
+    g = g';
 
 
     % We iterate through all voxels to solve the problem
     for i=1:param.NbrVoxels,
         
         % Initialisation of the regularisation parameter for the
         % considered voxel, in two steps (a,b):
 
         % a. Wavelet decomposition of the time course of the voxel
         % of interest. 
 
         
-        coeff=cconv(TCN(:,i),g);
+        coef=cconv(TCN(:,i),g);
 
 
         % c. Median absolute deviation (sum of absolute valued
         % distances of coefficients from the mean)
         % From the Matlab page on wavelet denoising:
         % "The median absolute deviation of the coefficients is a
         % robust estimate of noise."
         % This is thus going to be our estimate of the noise level
         % for the considered voxel time course
         param.LambdaTemp(i) = mad(coef,1)*param.LambdaTempCoef;
         
         % Now that we have estimated our initial lambda 
         % (regularisation parameter), Temporal_TA performs the 
         % computations themselves for the considered time course
         % TCN(i,:) of voxel i
 
         [TC_OUT(:,i),paramOUT] = TA_Temporal_OneTimeCourse(TCN(:,i),i,param);
 
         
         % Takes the final estimates of regularisation parameter and
         % 'effective noise' for voxel i, and sroes them in the
         % param structure
         param.LambdaTempFin(i) = paramOUT.LambdasTempFin;
         param.NoiseEstimateFin(i) = paramOUT.NoiseEstimateFin;  
     end  
 end