In the field of translation, there are inevitably going to be phrases and meanings in a Source Text that will not be able to be conveyed (at least not with complete fidelity) in the Target Text…that’s just the way it is. Even the best translators in the world will encounter fragments of text that simply are too peculiar to be able to be reproduced in a TT without translation loss, and given the unavoidability of translation loss, ways have been developed to offset it. This is what is referred to as compensation in the translation field.
It is the responsibility of individual translators to carefully balance and measure compromise and compensation in a TT they are producing. Too much compromise, and the TT will feel hollow; too much compensation, and it will be excessively drenched in the culture of the Target Language. There are four different types of compensation to be used.
Compensation in kind is basically the insertion of a a particular textual effect into the TT when the effect in the ST cannot be exactly reproduced; for example, English does not express gender of definite articles as Spanish does (el gato, la gata), and therefore a translator would need to compensate in kind by making a reference somewhere that makes this point clear.
Compensation in place is the shifting of a given textual effect from the exact place it is to be found in the ST to a different place (earlier or later) in the TT.
Compensation by merging is the practice of condensing a relatively extensive chunk of ST into a relatively shorter bit of TT. Often times, one language can express an identical or similar idea in far fewer words than is possible in the other language, such as often happens between English and Spanish.
In practice, the borders between these different types of compensation are murky, and often times a translator will make a decision that implies a little of all of them.