Parts of speech

The data shows two clear examples of CONVERSION — a change of word class with no change in the form of the word.

The first is 'google'. The table records it first as a noun (the name of the search engine, from 1998) and then as a verb whose use rises from around 2002 ('to google something'). No affix has been added: the noun and the verb are identical in form. This is conversion (also called functional shift or zero-derivation), and specifically a noun-to-verb conversion. 'Text' shows the same pattern even more sharply: the noun 'a text' (a message) acquires the verb 'to text' from around 2000, again with no change of form.

The second is 'friend', a long-established noun that the data shows gaining a verb use from around 2005 ('to friend someone' on a social network). Once more there is no affix — the verb is simply the noun used in a verb slot — so this too is noun-to-verb conversion.

These examples reveal that English readily forms new words by REUSING existing ones in a new class rather than always coining or borrowing. Conversion is highly PRODUCTIVE in present-day English: because the language has lost most of its inflections, a bare word can slot into a noun or verb position without any grammatical obstacle, and new technology supplies a steady stream of nouns (google, text, friend in its online sense) that speakers immediately convert to verbs because they need a verb for the action. Importantly, these new verbings are not 'errors' or 'not real words'; conversion is a long-standing, systematic process — Shakespeare himself converted freely — so the data evidences a normal mechanism of word-formation, not a decline in English.

Read correctly, this is a part-of-speech n-gram: it plots the word 'text' ONLY when tagged as a particular class, so 'text_NOUN' and 'text_VERB' appear as separate lines. The horizontal axis shows time (1950-2020) and the vertical axis shows relative frequency — the word-as-that-class as a proportion of all words in the printed sample. It is essential to read these axes and the tags first, because the graph tracks one word's class DISTRIBUTION, not a single word-count.

The trend is clear. The data shows 'text_NOUN' present throughout from 1950 and rising gently, while 'text_VERB' is near-zero until around 2000 and then rises steeply, narrowing the gap with the noun line by 2020.

This pattern is graphical evidence for CONVERSION — a part-of-speech change. The noun 'text' (a written message) has acquired a productive verb use, 'to text someone', and the sharp post-2000 rise tracks the spread of mobile-phone text-messaging. In other words, a word that was overwhelmingly a noun has shifted its class balance toward far greater verb use over roughly two decades — a word-class change we can date precisely from the data, driven by technology.

Two cautions strengthen the interpretation. First, part-of-speech tagging is done by SOFTWARE and is not perfect, so a novel verb use can be mis-tagged, especially early on. Second, the corpus is built from PRINTED, PUBLISHED books, which under-represent speech and informal messaging — so the verb 'to text' was almost certainly common in real use slightly before the printed record shows it. The graph evidences a pattern and a clear direction of change, not a precise statistic to quote as fact; used alongside a word-table, it is strong evidence of noun-to-verb conversion.

Read together, these two sources evidence one dominant mechanism of present-day word-class change: CONVERSION — the shift of a word from one part of speech to another with no change of form. The data shows conversion working in both directions (noun-to-verb and verb-to-noun) and reveals that English now forms a great deal of its new vocabulary by reusing existing words in new classes rather than only by borrowing or affixation. Organising the analysis by this issue, and drawing on both sources, lets each corroborate the other.

The clearest pattern is NOUN-TO-VERB conversion, and Sources A and B speak directly to it. Source A's word-table shows 'google' with around 0% verb use in 1995 but a substantial verb share by 2015 — a former proper-noun (a brand name) becoming an everyday verb, 'to google'. 'Friend', shown as roughly 100% noun in 1995, gains a noticeable verb share by 2015 ('to friend someone' online). At the morphological level the crucial point is that NO affix has been added: the verb is the noun used in a verb slot, which is precisely conversion (zero-derivation). Source B then supplies the diachronic detail for one such word: read with its tags and axes in mind (time against the relative frequency of 'text' as each class), it shows 'text_VERB' near-zero until around 2000 and then rising sharply, while 'text_NOUN' rises only gently. Synthesising the two, Source A captures the class SPLIT shifting across dates and Source B captures the same shift as a dated curve — both evidence the noun 'text' acquiring a productive verb use through conversion.

The reverse direction, VERB-TO-NOUN conversion, is also present in Source A: 'a read' and 'a fail' show rising NOUN shares of words that were verbs ('to read', 'to fail'). Again no affix is involved — the verb is simply used in a noun slot, often after an article ('that was a good read', 'the launch was a fail'). 'Gift', whose verb share rises from around 2005, shows conversion the other way (noun to verb, 'to gift'). The data therefore evidences conversion as a TWO-WAY, highly productive process, not a single one-directional drift.

A careful answer distinguishes a genuine NEW class from a mere rise in an existing use, and notes the data's limits. Some of these words may have had occasional verb or noun uses earlier than the table's first date; the figures track the BALANCE of uses, and a 'first verb use' in a printed corpus is the earliest surviving attestation, not necessarily the true first use in speech. The percentages also come from automated POS-tagging on a printed sample, so they carry tagger error and a written-language bias. These caveats do not weaken the overall pattern — they sharpen it.

Synthesising both sources gives a coherent account of word-formation. English forms much of its new vocabulary by CONVERSION because the language has lost most of its inflections, so a bare word can move freely between classes; technology and social media then supply a stream of new nouns (google, text, friend in its online sense) that are immediately converted to verbs. Far from being 'errors', these conversions continue a process as old as Shakespeare. The data thus shows the parts of speech of individual words shifting in neutral, systematic and highly productive ways — exactly the picture of word-class change the topic surveys.

These three sources, read together, evidence a key principle of word-class change: different parts of speech change at DIFFERENT RATES and for DIFFERENT REASONS, depending above all on whether they belong to a CLOSED class (which changes slowly and significantly) or an OPEN class (which changes rapidly and readily). Source A shows slow but meaningful change in the closed class of modal verbs; Sources B and C show rapid change and growth in the open classes of adverbs and verbs/nouns. Organising by this issue, rather than source by source, lets the sources illuminate one another.

The CLOSED-CLASS change is the modal-verb system in Source A. Modal verbs (shall, will, can, may, must, ought, should, would, could, might) are a closed class — genuinely new modals essentially never appear, so any shift within them is notable. The data shows certain modals declining across 1960-2010: 'shall', 'must' and 'ought (to)' fall markedly, while the so-called SEMI-MODALS — 'have to', 'going to', 'need to' — rise to take over their work. This is not a loss of the ability to express obligation or futurity; it is a shift WITHIN the system, the blunt modals giving way to softer, more periphrastic alternatives. The driver most often named is COLLOQUIALISATION and DEMOCRATISATION: language becoming less formal and less overtly authoritarian, so 'you must' yields to 'you need to' or 'you have to'. Because modals are a closed class, this change is slow and gradual, which is exactly what the gentle curves of the data show.

The OPEN-CLASS change is faster and more visible. Source B shows INTENSIFIERS — a sub-set of the open adverb class — turning over rapidly: 'really', 'so' and 'totally' rise across the late twentieth century while older intensifiers fade. Intensifiers are famously faddish because they suffer SEMANTIC BLEACHING: as an intensifier is overused it loses force, so speakers reach for a fresher one to sound emphatic. This fashion-driven churn is precisely what an open class permits and is a different kind of change from the slow modal shift — driven by fashion and bleaching, not by formality. Source C then shows the open classes of nouns and verbs GROWING through conversion: nouns such as 'google', 'friend' and 'message' gain verb uses after 2000, driven by technology and social media. Open classes readily gain members, so new verbings appear constantly.

Synthesising the three sources produces a coherent picture. The closed modal class (Source A) changes SLOWLY but meaningfully, its members redistributing toward semi-modals under colloquialisation. The open adverb class (Source B) changes RAPIDLY through fashion and semantic bleaching. The open noun/verb classes (Source C) GROW through productive conversion under technological pressure. These are not contradictory: they show that the RATE and CAUSE of change depend on the nature of the class. A measured answer also notes that all this data is drawn from printed, published corpora and (for Source C) automated tagging, so it lags and under-represents speech — but the central pattern is robust. The essence of word-class change is therefore not a single uniform drift but a system of classes moving at different speeds, for different reasons, all of it neutral and systematic rather than a sign of decline.

The statement makes two claims that need separating and testing: that CONVERSION is the most important way parts of speech change, and that TECHNOLOGY is its driver. The sources show that conversion is indeed a major and highly visible mechanism (Source A), and that technology is one real driver of it — but they also show that word classes change in OTHER important ways and for other reasons (Sources B and C). A balanced, evaluative answer therefore agrees the statement captures a central truth while qualifying it as too narrow.

THE CASE FOR THE STATEMENT: CONVERSION AND TECHNOLOGY. Source A strongly supports the first claim. It shows nouns 'google', 'text' and 'friend' gaining verb uses since 2000 — noun-to-verb CONVERSION (zero-derivation), in which the word changes class with no added affix. It also shows verb-to-noun conversion ('a read', 'a fail'), confirming that conversion is a two-way, highly productive process. The dating (post-2000) and the words themselves (google, text, friend in its online sense) point clearly to TECHNOLOGY and social media as a driver: new digital nouns appear and are immediately converted to verbs because speakers need a verb for the action. So far the statement holds: conversion is prominent and technology genuinely fuels much of it. Conversion is also favoured by a deeper structural fact — English has lost most of its inflections, so a bare word can move between classes freely — which is why technology's new nouns convert so easily.

THE CASE AGAINST THE STATEMENT'S NARROWNESS. Yet Sources B and C show change that is NOT conversion and NOT technology-driven. Source B evidences change within the CLOSED class of modal verbs: 'shall' and 'must' decline across 1960-2010 while the semi-modals 'have to' and 'going to' rise. This is a redistribution within a grammatical class, driven by COLLOQUIALISATION and DEMOCRATISATION (language becoming less formal and less authoritarian) — nothing to do with conversion or technology, and arguably MORE significant precisely because closed-class change is rare and slow. Source C evidences change in the OPEN adverb class: intensifiers 'really' and 'so' rise while older ones fade, driven by FASHION and SEMANTIC BLEACHING (an intensifier loses force through overuse, so a fresher one replaces it). Again this is neither conversion nor technological. These sources show that parts of speech change through several distinct mechanisms with several distinct drivers.

EVALUATION. The statement is therefore partly right but overstated. First, conversion is undeniably a MAJOR and conspicuous mechanism, and technology is a real driver of it (Source A) — the statement's instinct is sound. Second, however, 'most important' is hard to defend: closed-class change (Source B) is rarer and arguably more linguistically significant, and intensifier turnover (Source C) is a long-running, fashion-driven process older than any modern technology. Third, even conversion is not solely technological — verb-to-noun conversions like 'a read' owe more to informal speech style than to technology, and conversion as a process predates computers by centuries (Shakespeare converted freely). Fourth, all the data is drawn from printed, published corpora (and Source C from automated tagging), so it lags speech and may understate how early informal changes truly begin — a methodological caution that bears on any claim about which change is 'most important'.

In sum, the sources support the statement's core insight — conversion is a powerful, productive, partly technology-driven mechanism of word-class change (Source A) — but they refute its exclusivity. Parts of speech also change through closed-class redistribution under colloquialisation (Source B) and open-class turnover under fashion and semantic bleaching (Source C). The most accurate conclusion is that word classes change through SEVERAL neutral, systematic mechanisms at different rates; conversion is the most visible and one of the most productive, but it is not the whole story, and technology is one driver among several.

Studying how parts of speech change over time requires DATA that records word class, and POS-tagged corpora and part-of-speech n-gram graphs are the two principal quantitative methods for this. The sources let us test both their power and their limits, and a balanced evaluation concludes that they are uniquely valuable for revealing large-scale, dated shifts in word class but must be read with a clear understanding of what they do — and do not — capture. Engaging with that methodological question is itself the AO4 demand of this task.

THE STRENGTHS OF THE POS N-GRAM METHOD. Source A illustrates what part-of-speech n-grams do uniquely well: because they count a word ONLY when tagged as a given class, they can split a single word into its parts of speech and track each separately. The graph shows 'text' as a NOUN present from 1950 and rising gently, while 'text' as a VERB is near-zero until around 2000 and then rises sharply. No single text could show this; the method makes a word's class DISTRIBUTION visible and dateable across decades, capturing the conversion of the noun 'text' into a productive verb almost as it happened. Its scale and the precision of the tags give it real analytical power for word-class change specifically.

THE LIMITS OF THE POS N-GRAM METHOD. Yet Source A also exposes weaknesses a strong answer must read critically. First, axis-reading is essential: the vertical axis shows RELATIVE frequency, not raw counts, so a rise reflects proportion, not absolute number — misreading this is the commonest error. Second, the tagging is done by SOFTWARE, and automated POS-taggers are imperfect: they can misclassify words, and novel or ambiguous uses (exactly the new conversions we care about) are the hardest cases to tag correctly, so early verb counts may be unreliable. Third, the graph shows a PATTERN and correlation, not cause — it tells us 'to text' rose but not, by itself, WHY. Fourth, the data is from PRINTED, PUBLISHED books, so it lags and under-represents speech and informal messaging, where new verbings often start. The method captures the written, tagged, public language, not the whole language.

THE STRENGTHS AND LIMITS OF THE POS WORD-TABLE. Source B, the POS-tagged word-table, shares both the power and the caveats. By giving the PERCENTAGE of uses as each class at two dates, it lets us measure how a word's class balance has shifted — strong, systematic evidence of conversion and of rising or falling class membership. But its own note is the crucial caveat: the corpus is built from DIGITISED PRINTED BOOKS and tagged AUTOMATICALLY by software. So the figures carry tagger error, a written-language bias, and the usual corpus skews (what was printed, survived and was digitised, plus possible OCR errors). A 'first verb use' is the earliest surviving WRITTEN, tagged attestation, not the true birth of the use in speech. The table is therefore strong evidence for the printed, tagged record but a weaker guide to spoken usage and exact origins.

SYNTHESIS AND EVALUATION. These methods are most useful when their limits are respected and when they are COMBINED — with each other and with close analysis of real text. An n-gram trend (Source A) gains explanatory force when paired with a word-table (Source B) that quantifies the class split, and both gain force when checked against an actual dated extract showing the word in use, where a human reader can verify the class the tagger assigned. This triangulation — exactly the synthesis of varied sources AO5 rewards — compensates for the blind spots of any single method: the n-gram supplies the dated curve, the table the precise class split, the text the verified human reading. It also guards against the central methodological trap of confusing correlation with cause: the data shows THAT word classes shift, but the analyst must supply WHY (technology, colloquialisation, fashion).

In conclusion, POS-tagged corpora and part-of-speech n-grams are genuinely valuable, even indispensable, methods for studying word-class change: they reveal large-scale, dated, quantifiable shifts in part of speech that no individual text could show. But they are methods with defined limits — they record the printed, public, automatically tagged language, show pattern not cause, depend on imperfect software tagging, and require careful axis- and date-reading. Used critically and in combination, and never treated as a source of exact authoritative statistics, they are powerful tools; used naively, they mislead. The mark of sound linguistic method is precisely this awareness of what the data can and cannot tell us.