A not-so-pretty solution is this:

\begin{question}{Q1}
Which row in the table is correct about a uniform circular motion:\\

\hspace{1cm}velocity {\hskip 4em} acceleration
\begin{choices}
\wrongchoice{changing {\hskip 4em} constant}
\wrongchoice{constant {\hskip 4em} constant}
\wrongchoice{constant {\hskip 4em} changing}
\correctchoice{changing {\hskip 4em} changing}
\end{choices}
\end{question}

This will shuffle but maintain alignment in a very fragile way.
It might be better for your students to read actual sentence that you have in mind. That will be as unambiguous as can be:

\begin{question}{Q2}
Far uniform circular motion\\
\begin{choices}
\wrongchoice{velocity changes and acceleration is constant.}
\wrongchoice{velocity is constant and acceleration is constant.}
\wrongchoice{velocity is constant and acceleration is constant.}
\correctchoice{velocity changes and acceleration changes.}
\end{choices}
\end{question}

My attempts to interleave the tabbing or tabular environment with the question environment were quickly frustrated and I decided this is just as good. Of course, there might be questions where sentences are not ideal. In those circumstances you can typeset the question in a table format first, label the columns and then ask for the correct column in a horizontal fashion:

\begin{question}{Q3}
Which row in the table is correct about a uniform circular motion:\\
\begin{tabular}{r|r|r}
& velocity & acceleration\\
\hline
A & changing & constant \\
B & constant & constant \\
C & constant & changing \\
D & changing & changing \\
\end{tabular}\\
\begin{choiceshoriz}
\wrongchoice{A}
\wrongchoice{B}
\wrongchoice{C}
\correctchoice{D}
\end{choiceshoriz}
\end{question}

There you go - three methods to choose from, nothing quite like what you had in mind ;-)

Thank you very much for the advices. Indeed I've implemented the 2nd and 3rd myself and couldn't be satisfied.

I first went with the 3rd method. It worked with choice-shuffling turned off, as the order of the choices is determined in the question body. Shuffling the choices simply breaks the correspondence. Alternatively I coded the choices as
\wrongchoice[A]{}\wrongchoice[B]{}\wrongchoice[C]{}\correctchoice[D]{}
and allowed shuffling, with choices displayed like [B][A][D][C] or [D][C][A][B] etc. Somehow weird on exam papers.

Right now I'm settled with the 2nd method. Cumbersome at times, this method has natural shuffling and I don't need to worry about breaking things.

Maybe the perfect solution is to create a new environment like "choicescustom" in the package source. I'm looking at the .sty file, but have no idea how to do it at this point.

Regarding third method, have you considered using numbers as opposed to letters? Rows in the table could be named 1,2,3 and 4, and answers cold be coded \wrongchoice{1} or \wrongchoice[1$^st$ row} etc.

This is a good one. :-)

Natural and proper with minimal redundancy.

Nikola Z. Guscic wrote:

Regarding third method, have you considered using numbers as opposed to letters? Rows in the table could be named 1,2,3 and 4, and answers cold be coded wrongchoice{1} or wrongchoice[1$^st$ row} etc.