The product vs process distinction
This dichotomy is not new. It originates from Widdowson's distinction between textual product vs. discourse product.
Widdowson, H. 2000. ‘The limitations of linguistics applied’. Applied Linguistics 21/1: 3-25.
I have argued that corpus linguistics provides us with the description of text, not discourse. Although textual findings may well alert us to possible discourse significance and send us back to their contextual source, such significance cannot be read off from the data. The factual data constitute evidence of the textual product: what evidence they might provide of the discourse process is a matter for further enquiry.
Here is a response from Stubbs to this criticism:
Stubbs, M. 2001. ‘Texts, corpora, and problems of interpretation: a response to Widdowson’. Applied Linguistics 22/2: 149-172.
Process and product
Widdowson repeatedly argues that corpus linguistics provides us with a description of text as product, not discourse as process (pp. 6, 9, 10). Since a text is a ‘static semantic patchwork’ (pp. 7, 17, 22), which has been taken out of its social context of inference and interpretation, we can study only ‘textual traces’ (pp. 7, 11, 21, 22) of discourse process.
This is perfectly true, though the problem is very widespread in empirical disciplines. Recognizing the problem obviously does not solve it, but it shows that corpus linguistics is trying to develop observational, empirical methods of studying meaning, which are open to the same tests as are applied in other disciplines. For example, consider the parallels between corpus linguistics and geology, which both assume a relation between process and product. By and large, the processes are invisible, and must be inferred from the products.
Geologists are interested in processes which are not directly observable, because they take place across vast periods of time. What is observable is individual rocks and geographical formations: these products are the observable traces of processes which have often taken place a long time in the past. They are highly variable, because any specific instance is due to the local environment. Nevertheless, these variable products are due to highly general processes of destruction (such as erosion) and construction (such as sedimentation) (Love 1991).
Corpus linguists are interested in processes which are not directly observable because they are instantiated across the language use of many different speakers and writers. What is directly observable is the individual products, such as utterances and word combinations. (In addition, repetitions of such patterns, across time, can be made observable if different occurrences are displayed by concordancers and other software: see above.) These individual word combinations are the observable traces of general patterns of collocation and colligation. They are highly variable due to local socio-linguistic contexts. Nevertheless, these variable products are due to highly general processes of probability and speaker expectation.
This dichotomy is not new. It originates from Widdowson's distinction between textual product vs. discourse product.
Widdowson, H. 2000. ‘The limitations of linguistics applied’. Applied Linguistics 21/1: 3-25.
I have argued that corpus linguistics provides us with the description of text, not discourse. Although textual findings may well alert us to possible discourse significance and send us back to their contextual source, such significance cannot be read off from the data. The factual data constitute evidence of the textual product: what evidence they might provide of the discourse process is a matter for further enquiry.
Here is a response from Stubbs to this criticism:
Stubbs, M. 2001. ‘Texts, corpora, and problems of interpretation: a response to Widdowson’. Applied Linguistics 22/2: 149-172.
Process and product
Widdowson repeatedly argues that corpus linguistics provides us with a description of text as product, not discourse as process (pp. 6, 9, 10). Since a text is a ‘static semantic patchwork’ (pp. 7, 17, 22), which has been taken out of its social context of inference and interpretation, we can study only ‘textual traces’ (pp. 7, 11, 21, 22) of discourse process.
This is perfectly true, though the problem is very widespread in empirical disciplines. Recognizing the problem obviously does not solve it, but it shows that corpus linguistics is trying to develop observational, empirical methods of studying meaning, which are open to the same tests as are applied in other disciplines. For example, consider the parallels between corpus linguistics and geology, which both assume a relation between process and product. By and large, the processes are invisible, and must be inferred from the products.
Geologists are interested in processes which are not directly observable, because they take place across vast periods of time. What is observable is individual rocks and geographical formations: these products are the observable traces of processes which have often taken place a long time in the past. They are highly variable, because any specific instance is due to the local environment. Nevertheless, these variable products are due to highly general processes of destruction (such as erosion) and construction (such as sedimentation) (Love 1991).
Corpus linguists are interested in processes which are not directly observable because they are instantiated across the language use of many different speakers and writers. What is directly observable is the individual products, such as utterances and word combinations. (In addition, repetitions of such patterns, across time, can be made observable if different occurrences are displayed by concordancers and other software: see above.) These individual word combinations are the observable traces of general patterns of collocation and colligation. They are highly variable due to local socio-linguistic contexts. Nevertheless, these variable products are due to highly general processes of probability and speaker expectation.