diff --git a/paper.md b/paper.md
index aff882a..208a29f 100755
--- a/paper.md
+++ b/paper.md
@@ -53,7 +53,7 @@ where $m$, $n$, and $f$ are the transmit, receive and (image) frame indices, $i$
 Rather than require the user to specify the full five dimensional tensor $\alpha_{ijknm}$, any subset of these dimensions may be singular to efficiently form e.g. a transmit by lateral weighting (apodization) scheme or a receiver by depth weighting scheme.
 Permutations of the input data e.g. $x_{nfm}$ or $\mathbf{p}_{kji}$ are handled via properties specifying the indices rather than by enforcing a convention, which may lead to costly memory operations.
 
-Engaging in research often involves a plethora of small choices in order to configure software to suit the needs of a particular scientific inquiry. While exposing the user is necessary for some fields of research, this forces all users to consider each aspect of the methodology. 
+Engaging in research often involves a plethora of small choices in order to configure software to suit the needs of a particular scientific inquiry. While exposing the user to these choices is necessary for some fields of research, mandating a suitable input forces all users to consider each aspect of the methodology. 
 In most cases, the vast majority of these choices are either ultimately inconsequential or there exists a clear optimal or, at a minumum, a clear _acceptable_ choice.
 For example, for finite-difference time-domain (FDTD) simulators, knowledge of a transducer's frequency bandwidth is sufficient to determine a reasonably sized temporal sampling interval that satisfies the Courant-Friedrichs-Lewy (CFL) stability criterion.
 For research regarding the stability and accuracy of such simulators, the sampling interval is a critical choice. In many other cases however, this aspect is irrelevant as long as the sampling interval is _sufficiently_ small.