A great deal is written about difficulties in construing legal texts. Much less effort has gone into identifying interpretive problems that result from spoken language. This paper does that, by discussing how our abilities to perceive and understand speech lead to misunderstandings in legal contexts. Specifically, there are numerous battles over what was actually said in recorded speech. These disagreements are often reflected in disputes over how the interaction should be transcribed. We discuss many such examples, and explain them in terms of well-studied phonetic phenomena. These include our difficulty in segmenting speech into words (we speak without using a spacebar), and, in English, the fact that unstressed vowels, and some consonants, are reduced to the point of being indistinguishable, or even inaudible. For purposes of exposition, we compare cases involving the misperception of recorded speech in legal contexts with the misperception of song lyrics. Finally, we discuss our lack of memory for both the exact words spoken, and for human voices with which we are not sufficiently familiar. Our failure to recall exact words creates serious problems for the legal system with respect to prosecuting false statements made verbally, and enforcing oral contracts.