The concept of phraseology has come to increasing prominence in linguistics over the last two decades, but remains somewhat notorious for meaning different things to different researchers. In attempting to resolve this complexity, it has become more or less conventional for commentators to draw a distinction between two broadly contrasting approaches to the subject. One approach may be described as ‘typological’, in that it aims to develop formal taxonomies of word combination types, and to use such taxonomies to address theoretical and applied questions about language learning and teaching. The other approach may be referred to as ‘probabilistic’, in that it centrally involves using computational methods to identify phraseological units in large corpora of attested language data purely on the basis of
frequency and/or statistical significance of co-occurrence.
The aim of my talk is to submit this bipartite distinction to a critical review and reassessment. I begin by arguing that if we evaluate these two approaches to phraseology along the purely methodological lines described above, the probabilistic approach will inevitably fare much better than its typological rival. I then go on to consider another (and currently almost entirely neglected) way of comparing the two approaches, which focuses more strongly on evaluating the ontological status of the term ‘phraseology’ itself within each approach. I argue that the probabilistic approach fares much less well from this perspective, as it raises a number of fundamental theoretical questions which probabilistically-oriented
researchers have so far failed to address satisfactorily. I will conclude by pointing out that there is a third (and again all too often overlooked) conceptual approach to phraseology, which may offer a solution to the theoretical dilemmas currently facing researchers working within the probabilistic paradigm.