《汉语与认知》课程教学资源（认知语言学专著）认知语言学的研究方法（英文版）

XIV Leonard Talmy its degree of conformity to the standard patterns of syntax and morphology in one's lan- guage-though this can vary substantially across individuals,especially as a consequence of training.Most of generative linguistics rests on an assumption of the reliability of such grammaticality judgments. Still under the condition of attending to them in isolation,some aspects of language offer only moderate rather than strong accessibility to introspection.Returning to mean- ing as the object of attention,the strong access to the meaning of a word,noted earlier, holds whether the word is composed of one or more morphemes.But in the latter case, the meaning of a bound closed-class morpheme-that is,of an affix or clitic-seems only moderately accessible.Thus,while the meaning of the whole word unresettable,as in This kind of trap is unresettable,is quite available to direct examination,the meanings of the affixes un-,re-,and-able seem less fully and immediately available without the aid of ana- lytic manipulations.However,this aspect of language may be quite sensitive to differences across individual speakers.Such a difference may have existed between my two main con- sultants for Atsugewi-a polysynthetic Hokan language of northern California-both with little Western schooling and with their native language unwritten.One consultant could provide the single multi-affixal verb best expressing a situation I had depicted,but could not identify any of the affixes within the verb,such as the particular "Cause prefix"repre- senting the specific cause of the main event.The other consultant could not only provide the best verb for the depicted situation,but could proceed to volunteer systematic vari- ations of cause within the situation and provide the corresponding series of verbs,each one with a different Cause prefix.The second consultant would seem to have had greater conscious access to the morphological structure of words in her language. A further moderate case is that of access to the different meanings associated with a particular morphemic shape.For example,a person asked to think of the various senses of the noun stock might come up with several,but scarcely all,of the following:'soup base, stored supply,'line of descendants,farm animals,financial instrument,rifle part,fra- grant flowered plant species.One could then ask,though,what happened to the remaining senses in the respondent's cognition.On hearing the word stock in contexts that evoked them,the respondent would no doubt have come up with those very senses.Apparently, our cognition is organized in a way that allows particular senses of a word to come into consciousness in the relevant contexts,but not as a full connected set under introspec- tion-though introspection does yield a few.The full set can again typically be achieved only with the aid of analytic procedures,like those giving rise to dictionaries through a kind of corpus research.A similar accessibility pattern seems to hold for a comparable as- pect of language:approximate synonyms in a lexicon.If asked to think of other words with roughly the same meaning as,say,tendency,a respondent might come up with a couple, but probably not all,of the following:inclination,leaning,disposition,proneness,propen- sity,proclivity.If we think of a certain concept,our cognition is apparently so organized as to present in consciousness one or two of the lexical forms that express it and,with introspection,to present several more such forms,but not to present the extended set.As before,the extended set can be achieved only with the aid of analytic procedures through another kind of corpus research,the kind that leads to a thesaurus

JB[v.20020404] Prn:12/04/2007; 9:50 F: HCP18FR.tex / p.4 (236-300)  Leonard Talmy its degree of conformity to the standard patterns of syntax and morphology in one’s lan￾guage – though this can vary substantially across individuals, especially as a consequence of training. Most of generative linguistics rests on an assumption of the reliability of such grammaticality judgments. Still under the condition of attending to them in isolation, some aspects of language offer only moderate rather than strong accessibility to introspection. Returning to mean￾ing as the object of attention, the strong access to the meaning of a word, noted earlier, holds whether the word is composed of one or more morphemes. But in the latter case, the meaning of a bound closed-class morpheme – that is, of an affix or clitic – seems only moderately accessible. Thus, while the meaning of the whole word unresettable, as in This kind of trap is unresettable, is quite available to direct examination, the meanings of the affixes un-, re-, and -able seem less fully and immediately available without the aid of ana￾lytic manipulations. However, this aspect of language may be quite sensitive to differences across individual speakers. Such a difference may have existed between my two main con￾sultants for Atsugewi – a polysynthetic Hokan language of northern California – both with little Western schooling and with their native language unwritten. One consultant could provide the single multi-affixal verb best expressing a situation I had depicted, but could not identify any of the affixes within the verb, such as the particular “Cause prefix” repre￾senting the specific cause of the main event. The other consultant could not only provide the best verb for the depicted situation, but could proceed to volunteer systematic vari￾ations of cause within the situation and provide the corresponding series of verbs, each one with a different Cause prefix. The second consultant would seem to have had greater conscious access to the morphological structure of words in her language. A further moderate case is that of access to the different meanings associated with a particular morphemic shape. For example, a person asked to think of the various senses of the noun stock might come up with several, but scarcely all, of the following: ‘soup base’, ‘stored supply’, ‘line of descendants’, ‘farm animals’, ‘financial instrument’, ‘rifle part’, ‘fra￾grant flowered plant species’. One could then ask, though, what happened to the remaining senses in the respondent’s cognition. On hearing the word stock in contexts that evoked them, the respondent would no doubt have come up with those very senses. Apparently, our cognition is organized in a way that allows particular senses of a word to come into consciousness in the relevant contexts, but not as a full connected set under introspec￾tion – though introspection does yield a few. The full set can again typically be achieved only with the aid of analytic procedures, like those giving rise to dictionaries through a kind of corpus research. A similar accessibility pattern seems to hold for a comparable as￾pect of language: approximate synonyms in a lexicon. If asked to think of other words with roughly the same meaning as, say, tendency, a respondent might come up with a couple, but probably not all, of the following: inclination, leaning, disposition, proneness, propen￾sity, proclivity. If we think of a certain concept, our cognition is apparently so organized as to present in consciousness one or two of the lexical forms that express it and, with introspection, to present several more such forms, but not to present the extended set. As before, the extended set can be achieved only with the aid of analytic procedures through another kind of corpus research, the kind that leads to a thesaurus

Foreword XV Finally,now somewhat less directly involved with meaning,the lexical category of a word-noun,verb,adjective,etc.-also seems moderately accessible to introspection.Our cognition seems structured to allow us some sense of lexical category,although individuals surely vary on this and,even the most natively gifted individual would need training or practice to be able to articulate this sense. As the last circumstance under the condition of attending to a linguistic entity in isolation,some aspects of language appear to remain poorly or not at all accessible to introspection,that is,are elusive or absent under direct attention.Often,the means for getting at such aspects is to combine introspection-employing it where its capacities work well-with analytic manipulation.In the area of linguistic meaning,while the overall meaning of a word is one of the aspects of language most accessible to direct attention,the specific semantic components that structurally comprise that meaning largely are poorly accessible.For example,one can reasonably well attend in isolation to the overall mean- ing of the preposition across in its locative sense,as in The board lay across the road.And, if asked to,one could probably come up with a vague sense of what seemed like one of its components of meaning,say,'perpendicularity of two axes.But the combination of semantic introspection with analytic manipulation can tease out no fewer than nine crite- rial components of meaning (see Talmy 2003,2006).The method here is to systematically alter individual elements(especially spatial relations)within the referent situation and use one's semantic introspection to see if the original sentence with across still applies to it, or if now a sentence with a different preposition would apply instead.The situational el- ements needed for across to apply must then correspond to components of its meaning. Thus,one can little directly discern the semantic components of a word's meaning through introspection,but one can use introspection-in fact,perhaps one must use it-where it functions appropriately,in the experiencing of the overall meaning of a word,as part of a procedure for ferreting out its semantic components. Another aspect of language that does not simply pop into awareness on introspection is that of syntactic principles and patterns.For example,if asked to consider the two sen- tences Whose dog did our cat bite?and Whose dog bit our cat?,an average English speaker would have little direct sense for what it is about the first sentence that(among other syn- tactic characteristics)requires the inclusion of the word did,the basic form of the verb bite, and the positioning of this verb at the sentence's end,while the second sentence requires an absence of did,the past-tense form of the verb,and the positioning of the verb within the sentence.Again,though,linguists have been able to tease out syntactic regularities governing facts like these by combining their introspective sense for the grammatical- ity of sentences with the analytic procedure of systematically altering the elements and arrangements within such sentences to uncover patterns. Finally,there is little if any direct conscious access to the cognitive processes that un- derlie any of the strongly viable operations of introspection.For example,although we can readily attend to the fact that we get an immediate sense of the meaning of the word bucket when we hear it,we cannot attend to the mental processing that led to that meaning coming to mind

JB[v.20020404] Prn:12/04/2007; 9:50 F: HCP18FR.tex / p.5 (300-374) Foreword  Finally, now somewhat less directly involved with meaning, the lexical category of a word – noun, verb, adjective, etc. – also seems moderately accessible to introspection. Our cognition seems structured to allow us some sense of lexical category, although individuals surely vary on this and, even the most natively gifted individual would need training or practice to be able to articulate this sense. As the last circumstance under the condition of attending to a linguistic entity in isolation, some aspects of language appear to remain poorly or not at all accessible to introspection, that is, are elusive or absent under direct attention. Often, the means for getting at such aspects is to combine introspection – employing it where its capacities work well – with analytic manipulation. In the area of linguistic meaning, while the overall meaning of a word is one of the aspects of language most accessible to direct attention, the specific semantic components that structurally comprise that meaning largely are poorly accessible. For example, one can reasonably well attend in isolation to the overall mean￾ing of the preposition across in its locative sense, as in The board lay across the road. And, if asked to, one could probably come up with a vague sense of what seemed like one of its components of meaning, say, ‘perpendicularity of two axes’. But the combination of semantic introspection with analytic manipulation can tease out no fewer than nine crite￾rial components of meaning (see Talmy 2003, 2006). The method here is to systematically alter individual elements (especially spatial relations) within the referent situation and use one’s semantic introspection to see if the original sentence with across still applies to it, or if now a sentence with a different preposition would apply instead. The situational el￾ements needed for across to apply must then correspond to components of its meaning. Thus, one can little directly discern the semantic components of a word’s meaning through introspection, but one can use introspection – in fact, perhaps one must use it – where it functions appropriately, in the experiencing of the overall meaning of a word, as part of a procedure for ferreting out its semantic components. Another aspect of language that does not simply pop into awareness on introspection is that of syntactic principles and patterns. For example, if asked to consider the two sen￾tences Whose dog did our cat bite? and Whose dog bit our cat?, an average English speaker would have little direct sense for what it is about the first sentence that (among other syn￾tactic characteristics) requires the inclusion of the word did, the basic form of the verb bite, and the positioning of this verb at the sentence’s end, while the second sentence requires an absence of did, the past-tense form of the verb, and the positioning of the verb within the sentence. Again, though, linguists have been able to tease out syntactic regularities governing facts like these by combining their introspective sense for the grammatical￾ity of sentences with the analytic procedure of systematically altering the elements and arrangements within such sentences to uncover patterns. Finally, there is little if any direct conscious access to the cognitive processes that un￾derlie any of the strongly viable operations of introspection. For example, although we can readily attend to the fact that we get an immediate sense of the meaning of the word bucket when we hear it, we cannot attend to the mental processing that led to that meaning coming to mind

XVI Leonard Talmy 2.2 Second condition of attending We turn now to the second condition of attending,that is,directing one's attention to var- ious aspects of language occurring while one is engaged as a speaker or listener in ongoing discourse (which could be extended to fluent writing or reading).The various aspects of language present in discourse again seem to range widely in their accessibility to such meta-attention.Thus,strongly accessible are the overall topic or subject matter of some portion of discourse and-though this may well vary across individuals-a sense for the degree of thematic and logical coherence present in the discourse.Also highly accessible is the specific conceptual content expressed by a speaker one is listening to,even down to a relatively fine-grained level if one listens attentively.Perhaps a bit lower in accessibility to meta-attention,though still high,is the conceptual content that one has in mind to express as one speaks,as well as the conceptual content one cues up for one's next turn at speaking while still listening to one's interlocutor. But certain other aspects oflanguage seem to afford only moderate,little,or no acces- sibility to direct attention during communication.Thus,during speech,the recombinant core system of language(see Talmy 2004)is accompanied by expressive subsystems that the speaker generally seems able to attend to only sporadically and moderately.Perhaps in order of decreasing accessibility,these include "vocal dynamics",my term for the gradi- ent subsystem in language that includes pitch,loudness,rate,precision,etc.;the speaker's gestures,including not only their unfolding forms but also their timing relative to the ongoing spoken track;and the speaker's facial expressions and body language. Deviations from well-formedness in speech can attract speaker or hearer attention and can even appear in attention with full salience if they exceed a certain "grace al- lowance",but those occurring within the grace allowance scarcely attract attention and are only moderately accessible to any attention directed at them.Such deviations in- clude contextually inappropriate forms,misplaced forms,non-optimal choice of words or constructions,grammatical or referential conflict across different sentence portions, self-corrections,incomplete constructions,pauses,inclusion of "uh"and "oh",restarts, interruptions by other speakers,and overlaps with other speakers.Also attracting little at- tention and perhaps highly affected by any attention directed at it is what a speaker's or hearer's gaze successively lights on,in the former case perhaps influencing what she next says and in the latter case often influenced by what the speaker has just said. Finally,seemingly inaccessible to conscious attention are the cognitive operations and processes going on that yield the production or the comprehension of speech. 2.3 Third condition of attending Without examining it in detail,the third condition ofattending-directing attention to the memory trace of various aspects of language that have just been manifested-also exhibits different degrees of accessibility.Thus,at the relatively high end of accessibility might be the ability to attend to the thematic topic and conceptual content,still in memory,of some immediately prior discourse.By contrast,the memory of,and hence one's ability to attend

JB[v.20020404] Prn:12/04/2007; 9:50 F: HCP18FR.tex / p.6 (374-420)  Leonard Talmy . Second condition of attending We turn now to the second condition of attending, that is, directing one’s attention to var￾ious aspects of language occurring while one is engaged as a speaker or listener in ongoing discourse (which could be extended to fluent writing or reading). The various aspects of language present in discourse again seem to range widely in their accessibility to such meta-attention. Thus, strongly accessible are the overall topic or subject matter of some portion of discourse and – though this may well vary across individuals – a sense for the degree of thematic and logical coherence present in the discourse. Also highly accessible is the specific conceptual content expressed by a speaker one is listening to, even down to a relatively fine-grained level if one listens attentively. Perhaps a bit lower in accessibility to meta-attention, though still high, is the conceptual content that one has in mind to express as one speaks, as well as the conceptual content one cues up for one’s next turn at speaking while still listening to one’s interlocutor. But certain other aspects of language seem to afford only moderate, little, or no acces￾sibility to direct attention during communication. Thus, during speech, the recombinant core system of language (see Talmy 2004) is accompanied by expressive subsystems that the speaker generally seems able to attend to only sporadically and moderately. Perhaps in order of decreasing accessibility, these include “vocal dynamics”, my term for the gradi￾ent subsystem in language that includes pitch, loudness, rate, precision, etc.; the speaker’s gestures, including not only their unfolding forms but also their timing relative to the ongoing spoken track; and the speaker’s facial expressions and body language. Deviations from well-formedness in speech can attract speaker or hearer attention and can even appear in attention with full salience if they exceed a certain “grace al￾lowance”, but those occurring within the grace allowance scarcely attract attention and are only moderately accessible to any attention directed at them. Such deviations in￾clude contextually inappropriate forms, misplaced forms, non-optimal choice of words or constructions, grammatical or referential conflict across different sentence portions, self-corrections, incomplete constructions, pauses, inclusion of “uh” and “oh”, restarts, interruptions by other speakers, and overlaps with other speakers. Also attracting little at￾tention and perhaps highly affected by any attention directed at it is what a speaker’s or hearer’s gaze successively lights on, in the former case perhaps influencing what she next says and in the latter case often influenced by what the speaker has just said. Finally, seemingly inaccessible to conscious attention are the cognitive operations and processes going on that yield the production or the comprehension of speech. . Third condition of attending Without examining it in detail, the third condition of attending – directing attention to the memory trace of various aspects of language that have just been manifested – also exhibits different degrees of accessibility. Thus, at the relatively high end of accessibility might be the ability to attend to the thematic topic and conceptual content, still in memory, of some immediately prior discourse. By contrast, the memory of, and hence one’s ability to attend

Foreword xvI to,the exact wording and phrasing that had just been used to represent such conceptual content might be elusive. 2.4 Profile of introspection A profile has just been sketched of the capacities and limitations that introspection has in accessing aspects of language.Where this profile has peaks and elevations,introspection may be the methodology of necessity or choice.But as its profile dips,other methodolo- gies become increasingly necessary to corroborate or fill in for the introspective findings. And where introspection lacks access altogether,any information about those aspects of language must come from other methodologies. In addition,researchers can feel a call to other methodologies to the extent that they have either or both of the following two concerns.The first is the concern that observa- tion can affect what is observed.For the present intra-cognitive issue,this is the concern that directing second-level attention at them might disturb the contents of first-level con- sciousness.The claimed gradient by which first-level contents range from being more stable to being more elusive either can be accepted and cause increasing concern toward the elusive end,or can be questioned even at the proposed stable end.Further,concern is likely to be greater for the second condition of attending than for the first,in part because some of the attention otherwise needed just to maintain adequacy of speaking or listening in a conversation might be diverted to the purpose of self-observation. The second concern is over the reliability of and efficacy of introspection in the first place.The reservation can be over the existence of consciousness itselfor,if that is granted, over whether the contents of consciousness can accurately reflect the unconscious mental processes felt to constitute the bulk of cognitive functioning. 3.Audio-and videographic analysis Our look at several other methodologies for their respective capacities and limitations can start with that of examining audiovisual recordings of speech events-that is,audio-and videographic analysis.This methodology has certain advantages in accessing aspects of language where introspection is limited.One such advantage is the ability to examine at leisure the online expressive accompaniments of the core speech track that were earlier cited as only of moderate to low accessibility to second-condition introspection.These include a speaker's vocal dynamics,gestures,facial expressions,body language,and gaze direction.Included as well is the exact timing of all these accompaniments,both alone and in relation to each other.Live recording can capture for study many aspects of the core speech track as well-also not readily attended to during the event-such as its deviations from well-formedness.In addition,what a speaker says at any given moment can be regarded as the result of numerous interacting factors,exerting their various calls on the speaker (see Talmy 2000b:Ch.6).These factors are sensitive to the immediate idea as well as the larger ideational complex needing expression,to assumptions about what the addressee already knows or should be informed of,to what needs emphasis or de-

JB[v.20020404] Prn:12/04/2007; 9:50 F: HCP18FR.tex / p.7 (420-481) Foreword  to, the exact wording and phrasing that had just been used to represent such conceptual content might be elusive. . Profile of introspection A profile has just been sketched of the capacities and limitations that introspection has in accessing aspects of language. Where this profile has peaks and elevations, introspection may be the methodology of necessity or choice. But as its profile dips, other methodolo￾gies become increasingly necessary to corroborate or fill in for the introspective findings. And where introspection lacks access altogether, any information about those aspects of language must come from other methodologies. In addition, researchers can feel a call to other methodologies to the extent that they have either or both of the following two concerns. The first is the concern that observa￾tion can affect what is observed. For the present intra-cognitive issue, this is the concern that directing second-level attention at them might disturb the contents of first-level con￾sciousness. The claimed gradient by which first-level contents range from being more stable to being more elusive either can be accepted and cause increasing concern toward the elusive end, or can be questioned even at the proposed stable end. Further, concern is likely to be greater for the second condition of attending than for the first, in part because some of the attention otherwise needed just to maintain adequacy of speaking or listening in a conversation might be diverted to the purpose of self-observation. The second concern is over the reliability of and efficacy of introspection in the first place. The reservation can be over the existence of consciousness itself or, if that is granted, over whether the contents of consciousness can accurately reflect the unconscious mental processes felt to constitute the bulk of cognitive functioning. . Audio- and videographic analysis Our look at several other methodologies for their respective capacities and limitations can start with that of examining audiovisual recordings of speech events – that is, audio- and videographic analysis. This methodology has certain advantages in accessing aspects of language where introspection is limited. One such advantage is the ability to examine at leisure the online expressive accompaniments of the core speech track that were earlier cited as only of moderate to low accessibility to second-condition introspection. These include a speaker’s vocal dynamics, gestures, facial expressions, body language, and gaze direction. Included as well is the exact timing of all these accompaniments, both alone and in relation to each other. Live recording can capture for study many aspects of the core speech track as well – also not readily attended to during the event – such as its deviations from well-formedness. In addition, what a speaker says at any given moment can be regarded as the result of numerous interacting factors, exerting their various calls on the speaker (see Talmy 2000b:Ch. 6). These factors are sensitive to the immediate idea as well as the larger ideational complex needing expression, to assumptions about what the addressee already knows or should be informed of, to what needs emphasis or de-

xvII Leonard Talmy emphasis,etc.;and yields selections as to morphemes and constructions.Introspection cannot readily access all these factors online because there are too many of them,because many are elusive,and because some of them may be out of awareness. Further,these factors appear to be triggered in their full complement in the cognition of the speaker only when he is engaged in what he understands to be an actual interlocutive event of speaking.Hence,audiovisual recordings register what people actually say during naturalistic speech,which cannot be replicated by the kinds of sentences often constructed out of context for first-condition introspection in isolation. Finally,longer stretches of live recording permit analysis of the kind of discourse func- tions central to functional linguistics,which,due to their length,are minimally accessible to the isolative introspection that excels at short linguistic forms. One main limitation to the methodology of working with spontaneous communica- tion-given that such communication is the result of many interacting cognitive factors- is that it does not lead directly to determining the lawful properties of any single factor by itself.For this,the more isolative,probing methods are generally needed.Perhaps a phys- ical analogy might be that,while one kind of study can characterize the complex pattern in which,say,a feather falls,another kind of study is needed to separate out and individ- ually characterize the contributing components,such as gravity,turbulence,friction,and buoyancy. 4.Corpus analysis Corpora,whose use constitutes another methodology,share with audiovisual recording a focus on naturalistically produced speech.But they typically capture only its segmen- tal text within their usually written format.Live recording thus has the advantage over corpora in capturing the expressive accompaniments of text and the timing of all these components.The consequent advantage of corpora,though,is to make a large quantity of textual discourse available to searches for particular phenomena of interest. This advantage is especially great for linguistic phenomena whose frequency of oc- currence or range of instantiations is the issue.Thus,corpus research can determine the frequency with which a particular expression is used in various contexts.This can be interpreted as a reflection of its unconsciously registered values along various cognitive parameters.These unconscious values are otherwise reflected in consciousness and avail- able to introspection only approximately-as a phenomenological sense of the expression's colloquiality and register.Searches through chronologically sequenced texts,in addition, can uncover gradual changes in the frequency-and therefore changes in the unconscious parametric values-of an expression,whereas such change is largely too slow for any di- rect introspection.Frequency assessments can also be made of a particular morpheme's co-occurrence with other morphemes.Once again,introspection could yield an approx- imate sense for the naturalness of such collocations,but not of their specific patterns of occurrance. Another special asset of corpus work is that it helps one map out the range of alterna- tive realizations of a linguistic entity.Examples of such ranges are the set of constructions

JB[v.20020404] Prn:12/04/2007; 9:50 F: HCP18FR.tex / p.8 (481-550)  Leonard Talmy emphasis, etc.; and yields selections as to morphemes and constructions. Introspection cannot readily access all these factors online because there are too many of them, because many are elusive, and because some of them may be out of awareness. Further, these factors appear to be triggered in their full complement in the cognition of the speaker only when he is engaged in what he understands to be an actual interlocutive event of speaking. Hence, audiovisual recordings register what people actually say during naturalistic speech, which cannot be replicated by the kinds of sentences often constructed out of context for first-condition introspection in isolation. Finally, longer stretches of live recording permit analysis of the kind of discourse func￾tions central to functional linguistics, which, due to their length, are minimally accessible to the isolative introspection that excels at short linguistic forms. One main limitation to the methodology of working with spontaneous communica￾tion – given that such communication is the result of many interacting cognitive factors – is that it does not lead directly to determining the lawful properties of any single factor by itself. For this, the more isolative, probing methods are generally needed. Perhaps a phys￾ical analogy might be that, while one kind of study can characterize the complex pattern in which, say, a feather falls, another kind of study is needed to separate out and individ￾ually characterize the contributing components, such as gravity, turbulence, friction, and buoyancy. . Corpus analysis Corpora, whose use constitutes another methodology, share with audiovisual recording a focus on naturalistically produced speech. But they typically capture only its segmen￾tal text within their usually written format. Live recording thus has the advantage over corpora in capturing the expressive accompaniments of text and the timing of all these components. The consequent advantage of corpora, though, is to make a large quantity of textual discourse available to searches for particular phenomena of interest. This advantage is especially great for linguistic phenomena whose frequency of oc￾currence or range of instantiations is the issue. Thus, corpus research can determine the frequency with which a particular expression is used in various contexts. This can be interpreted as a reflection of its unconsciously registered values along various cognitive parameters. These unconscious values are otherwise reflected in consciousness and avail￾able to introspection only approximately – as a phenomenological sense of the expression’s colloquiality and register. Searches through chronologically sequenced texts, in addition, can uncover gradual changes in the frequency – and therefore changes in the unconscious parametric values – of an expression, whereas such change is largely too slow for any di￾rect introspection. Frequency assessments can also be made of a particular morpheme’s co-occurrence with other morphemes. Once again, introspection could yield an approx￾imate sense for the naturalness of such collocations, but not of their specific patterns of occurrance. Another special asset of corpus work is that it helps one map out the range of alterna￾tive realizations of a linguistic entity. Examples of such ranges are the set of constructions